The present invention relates to a technology for identifying a vehicle on a road using information of an image photographed by a camera.
Advanced driver assistance systems (ADAS) such as automatic braking and ACC (Adaptive Cruise Control) have been marketed in a general market. Research and development of an autonomous driving system have also been under way, and a test of an autonomous vehicle on public roads has been performed.
Among vehicles and autonomous vehicles with the advanced driving assistance systems installed therein, there is a vehicle for which, by using a photographed image obtained by photographing around the vehicle by a vehicle-mounted camera, image recognition of a surrounding vehicle that is present around the vehicle is performed and the surrounding vehicle is detected.
As a method of detecting the surrounding vehicle using an image recognition process, there is a method of storing a large amount of image information of various vehicles as reference images in a database in advance and performing a pattern matching process between the photographed image and the reference images.
As the pattern matching process, there is a method of computing HoG (Histograms of Oriented Gradients) feature amounts using the reference images and the photographed image and determining whether or not the vehicle is photographed in the photographed image by an SVM (Support Vector Machine). As the pattern matching process, there is also a method of detecting, by a CNN (Convolutional Neural Network), a photographed image region similar to a feature amount map indicating a feature of the vehicle.
As the reference images in the database, a correct answer image being image information obtained by photographing the vehicle and an incorrect answer image being image information in which the vehicle is not photographed are employed.
By detecting, by the image recognition process, the image region of the vehicle included in the image information obtained by the photographing, the position of the surrounding vehicle that is present around the vehicle can also be detected. By utilizing a result of this detection, the advanced driver assistance systems such as the automatic braking and the ACC and a process of determining the travel route of an autonomous vehicle can be implemented.
An amount of computation for the image recognition process is large. Therefore, if a low-cost image processing apparatus is used, an image processing speed is reduced, so that application to systems, such as the advanced driver assistance systems and the autonomous driving system, which demand real-time performance, will become difficult.
Further, depending on a traveling environment, there is a case where the surrounding vehicle is not detected even if the surrounding vehicle is photographed in the photographed image, and on the contrary, there is a case where the surrounding vehicle is erroneously detected even if the surrounding vehicle is not photographed. In particular, a direction where the surrounding vehicle is photographed differs according to a relative position between the vehicle-mounted camera of the vehicle and the surrounding vehicle. Therefore, image information of the surrounding vehicle photographed in the photographed image greatly differs, so that a detection rate of the surrounding vehicle in the image recognition process will be reduced.
As mentioned above, the detection accuracy of the surrounding vehicle is reduced due to the traveling environment and the environment of photographing. Thus, the ACC to reduce a burden of the driver at a time of a traffic jam and the automatic braking as an assistance when a driver has made a mistake in driving are commercialized. However, the autonomous vehicle has not been commercialized yet. An image recognition technology with higher detection accuracy of a surrounding vehicle and a less computation amount than ever before is desired in order to enhance safety of an automobile, as well as to market the autonomous vehicle.
Patent Literature 1 describes that a degree of skew, which is a relative angle between a vehicle and a surrounding vehicle, is determined, an image recognition algorithm for the surrounding vehicle is switched according to the magnitude of the degree of the skew, and the surrounding vehicle is detected using a photographed image. Specifically, according to Patent Literature 1, when the degree of the skew is low, the surrounding vehicle is detected by the pattern matching process, and when the degree of the skew is high, the surrounding vehicle is detected by an optical flow process. That is, according to Patent Literature 1, when the degree of the skew between the vehicle and the surrounding vehicle is high, so that a reference image photographed from a front side and the photographed image greatly differ, the surrounding vehicle is detected by the optical flow process without using the reference image.
Patent Literature 1: JP 2013-161202 A
In the optical flow process, identification of the type of the surrounding vehicle and identification between the vehicle and a moving object other than the vehicle become difficult. Therefore, according to the technology described in Patent Literature 1, when the degree of the skew is high, the type and so on of the surrounding vehicle cannot be recognized.
An object of the present invention is to allow detection of a surrounding vehicle at high speed and with high accuracy using information of an image photographed by a camera.
A vehicle determination apparatus according to the present invention may include:
a direction identification unit to identify a traveling direction in which a surrounding vehicle travels in a partial region of a region indicated by image information obtained by photographing by a camera;
a feature amount acquisition unit to acquire an image feature amount computed from a reference image corresponding to the traveling direction identified by the direction identification unit; and
a vehicle determination unit to compute an image feature amount of the image information of the partial region and compare the computed image feature amount with the image feature amount acquired by the feature amount acquisition unit, thereby determining whether or not the surrounding vehicle is present in the partial region.
In the present invention, using the reference image corresponding to the traveling direction of the surrounding vehicle in the partial region to be processed, it is determined whether or not the surrounding vehicle is present in the partial region. With this arrangement, the process can be performed at high speed by reducing the number of reference images to be used. Further, since an appropriate reference image is used, the surrounding vehicle can be detected with high accuracy.
A configuration of a vehicle determination apparatus 10 according to a first embodiment will be described with reference to
The vehicle determination apparatus 10 is a computer that is mounted on a vehicle 100. The vehicle determination apparatus 10 is connected to a vehicle control unit 32 mounted on the vehicle 100 via a vehicle control interface 321.
The vehicle determination apparatus 10 includes a processor 11, a storage device 12, and a camera interface 13. The processor 11 is connected to the other hardware via signal lines and controls these other hardware.
The processor 11 is an IC (Integrated Circuit) to perform processing. As a specific example, the processor 11 is a CPU (Central Processing Unit), a DSP (Digital Signal Processor), or a GPU (Graphics Processing Unit).
The storage device 12 includes a memory 121 and a storage 122. As a specific example, the memory 121 is a RAM (Random Access Memory). As a specific example, the storage 122 is a ROM (Read Only Memory). The storage 122 may be an HDD (Hard Disk Drive). Alternatively, the storage 122 may be a portable storage medium such as an SD (Secure Digital) memory card, a CF (Compact Flash), a NAND flash, a flexible disk, an optical disk, a compact disk, a blue ray (registered trade mark) disk, or a DVD.
The camera interface 13 is a device to connect a camera 31 mounted on the vehicle 100 for photographing around the vehicle 100. As a specific example, the camera interface 13 is an interface substrate to connect the camera 31. By changing the camera interface 13 according to an output connecter, various types of the camera 31 can be connected to the vehicle determination apparatus 10.
The vehicle determination apparatus 10 includes a direction identification unit 21, a feature acquisition unit 22, and a vehicle determination unit 23, as functional components. A function of each unit of the direction identification unit 21, the feature acquisition unit 22, and the vehicle determination unit 23 is implemented by software.
A program to implement the function of each unit of the vehicle determination apparatus 10 is stored in the storage 122 of the storage device 12. This program is loaded into the memory 121 by the processor 11 and is executed by the processor 11. This implements the function of each unit of the vehicle determination apparatus 10. Reference feature amounts 41 to be used by the feature acquisition unit 22 are stored in the storage 122 of the storage device 12. The reference feature amounts 41 may be the ones stored in an external server.
Information, data, signal values, and variable values indicating results of processes of the functions of the respective units that are implemented by the processor 11 are stored in the memory 121 or a register or a cache memory in the processor 11. In the following description, it is assumed that the information, the data, the signal values, and the variable values indicating the results of the processes of the functions of the respective units that are implemented by the processor 11 are stored in the memory 121.
The program to implement each function to be implemented by the processor 11 has been assumed to be stored in the storage device 12. This program, however, may be stored in a portable storage medium such as a magnetic disk, a flexible disk, an optical disk, a compact disk, a blue ray (registered trademark) disk, or a DVD.
Only one processor 11 has been illustrated in
The vehicle control unit 32 is a computer that is mounted on the vehicle 100.
The vehicle control unit 32 includes the vehicle control interface 321, a sensor ECU (Electric Control Unit) 322, and a vehicle ECU 323.
The vehicle control interface 321 is an interface for being connected to a different apparatus such as the vehicle determination apparatus 10.
The sensor ECU 322 is a device to which various sensors such as a speed sensor, an acceleration sensor, a gyro sensor, a laser sensor, and a milliwave sensor are connected and which obtains information from the various sensors.
The vehicle ECU 323 is connected to various control devices for controlling the vehicle 100, such as a brake, an accelerator, and a steering wheel of the vehicle 100. The vehicle ECU 323 controls the various control devices based on the information acquired from the various sensors by the sensor ECU 322 and information transmitted from an external apparatus connected via the vehicle control interface 321, thereby controlling operations of the vehicle 100.
The operations of the vehicle determination apparatus 10 according to the first embodiment will be described with reference to
The operations of the vehicle determination apparatus 10 according to the first embodiment correspond to a vehicle determination method according to the first embodiment. The operations of the vehicle determination apparatus 10 according to the first embodiment correspond to a vehicle determination program procedure according to the first embodiment.
In the following description, the description will be given, using a case of driving on the left side of a road, as an example. A case of driving on the right side of the road may be considered to be the one in which the right side and the left side are reversed.
The operations of the vehicle determination apparatus 10 according to the first embodiment will be outlined with reference to
Referring to
An image greatly differs between when each surrounding vehicle 200 has been photographed from the front and when the surrounding vehicle 200 has been photographed from the rear. That is, the image of the surrounding vehicle 200 to be obtained when photographed by the camera 31 differs according to the traveling direction of the surrounding vehicle 200. When the image differs, a feature amount to be computed from that image also differs.
Then, the vehicle determination apparatus 10 stores the reference feature amount 41 for each photographing direction in which each surrounding vehicle 200 is photographed. Then, with respect to each partial region 61 of a region indicated by the image information 42, the vehicle determination apparatus 10 identifies the traveling direction of the surrounding vehicle 200 in the partial region 61 to be processed and determines whether or not the surrounding vehicle 200 is present in the partial region 61 to be processed, using the reference feature amount 41 corresponding to the identified traveling direction.
The reference feature amount 41 according to the first embodiment will be described with reference to
The reference feature amount 41 is computed by using reference images corresponding to a target photographing direction. Referring to
The reference feature amount 41 corresponding to each photographing direction is stored in the storage 122, being associated with each photographing direction as illustrated in
Herein, the reference feature amount 41A specialized in the image of the vehicle rear portion and the reference feature amount 41B specialized in the image of the vehicle front portion have been computed.
When the reference feature amount 41A is computed, an image on a left oblique rear portion of each of the vehicles, an image on a right oblique rear portion of the vehicle, an image on a left side portion of the vehicle, or an image on a right side portion of the vehicle as well as an image just behind the vehicle may be included. When the reference feature amount 41B is computed, an image on a right oblique front portion of the vehicle or an image on the right side portion of the vehicle as well as an image in front of the vehicle may be included. Herein, the driving on the left side of the road is assumed. Thus, the opposite lane is located on the right side of the vehicle 100. Therefore, when the reference feature amount 41B is computed, no images on the left sides of the vehicles are used.
The operations of the vehicle determination apparatus 10 according to the first embodiment will be described in detail, with reference to
In the first embodiment, processes illustrated in
In a photographed image acquisition process in step S101, the direction identification unit 21 acquires, via the camera interface 13, image information 42 in which the front of a vehicle 100 has been photographed by a camera 31. The direction identification unit 21 writes, into the memory 121, the image information 42 that has been acquired.
In a region selection process in step S102, the direction identification unit 21 selects a partial region 61 of a region indicated by the image information 42 acquired in step S101.
A specific description will be given with reference to
In a direction identification process in step S103, the direction identification unit 21 identifies a traveling direction in which a surrounding vehicle 200 travels in the partial region 61 selected in step S102.
Specifically, the direction identification unit 21 reads one of the partial regions 61 from the memory 121 and determines whether the partial region 61 that has been read is a region of a parallel lane 51 or a region of an opposite lane 52. Then, if the partial region 61 is the region of the parallel lane 51, the direction identification unit 21 identifies that the traveling direction of the surrounding vehicle 200 is the same direction as that of the vehicle 100. If the partial region 61 is the region of the opposite lane 52, the direction identification unit 21 identifies that the traveling direction of the surrounding vehicle 200 is the opposite direction to that of the vehicle 100. The direction identification unit 21 writes direction information 43 indicating the identified traveling direction into the memory 121.
A method of identifying whether the partial region 61 is the region of the parallel lane 51 or the region of the opposite lane 52 will be described with reference to
In the case of driving on the left side of the road, the parallel lane 51 is on the left side of the image information 42, and the opposite lane 52 is on the right side of the image information 42 in the camera 31 for photographing the front of the vehicle 100. Then, the direction identification unit 21 sets, as the parallel lane 51, a region from the left end to the center position or a position on a slightly right side of the center position the image information 42, and sets, as the opposite lane 52, a region from the right end to the center position or a position on a slightly left side of the center position of the image information 42.
Referring to
The direction identification unit 21 determines whether the partial region 61 is the region of the parallel lane 51 or the region of the opposite lane 52 according to which of the region of the parallel lane 51 and the region of the opposite lane 52 the partial region 61 includes. If the partial region 61 includes both of the region of the parallel lane 51 and the region of the opposite lane 52, the direction identification unit 21 determines that the partial region 61 is both of the regions of the parallel lane 51 and the opposite lane 52.
That is, in the method described with reference to
A method of narrowing the overlapping region will be described with reference to
It is assumed that a surrounding vehicle 200 that travels in the parallel lane 51 was detected by the vehicle determination unit 23 in the process that will be described later. In this case, a region where the surrounding vehicle 200 was detected is presumed to be in the parallel lane 51 rather than in the opposite lane 52. Then, when the surrounding vehicle 200 that travels in the parallel lane 51 was detected, the direction identification unit 21 sets, as the opposite lane 52, a region from the right end of image information 42 to the right end of the region where the surrounding vehicle 200 was detected. This reduces the region of the opposite lane 52 and narrows the overlapping region.
Similarly, when a surrounding vehicle 200 that travels in the opposite lane 52 was detected, the direction identification unit 21 sets, as the parallel lane 51, a region from the left end of image information 42 to the left end of a region where the surrounding vehicle 200 was detected. This reduces the region of the parallel lane 51 and narrows the overlapping region.
That is, in the method described with reference to
In a feature amount acquisition process in step S104, the feature acquisition unit 22 acquires a reference feature amount 41 corresponding to the traveling direction of the surrounding vehicle 200 identified in step S103.
Specifically, if the traveling direction of the surrounding vehicle 200 has been identified to be the same direction as that of the vehicle 100, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the same direction as that of the vehicle 100. That is, the feature acquisition unit 22 reads the reference feature amount 41A for the vehicle rear portion from the storage 122. On the other hand, if the traveling direction of the surrounding vehicle 200 has been identified to be the opposite direction to that of the vehicle 100, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the opposite direction to that of the vehicle 100. That is, the feature acquisition unit 22 reads the reference feature amount 41B for the vehicle front portion from the storage 122. Then, the feature acquisition unit 22 writes, into the memory 121, the reference feature amount 41 that has been read.
In a feature amount computation process in step S105, the vehicle determination unit 23 computes an image feature amount that is a feature amount of the partial region 61 selected in step S102.
Specifically, in the first embodiment, the vehicle determination unit 23 computes an HOG feature amount of the partial region 61, as the image feature amount. First, the vehicle determination unit 23 computes the luminance gradient intensity and the luminance gradient direction of each cell 62 constituting the partial region 61. Then, the vehicle determination unit 23 divides the computed luminance gradient direction into nine directions and generates a histogram of the luminance gradient intensity for each direction. Finally, the vehicle determination unit 23 sets all histograms of the respective cells 62, as the image feature amount of the partial region 61. The vehicle determination unit 23 writes, into the memory 121, the image feature amount computed.
Herein, the image feature amount of the partial region 61 becomes the image feature amount of a total of 324 (=9×36) dimensions because each cell 62 has the luminance gradient intensities of nine directions and the number of the cells 62 constituting each partial region 61 is 36.
The reference feature amount 41 stored in the storage 122 has been computed similarly to the image feature amount, using the reference images classified according to the vehicle direction. The reference feature amount 41 computed in advance may be stored, or may be arranged to be updated based on the image information 42.
Herein, the HOG feature amount has been computed as each of the image feature amount and the reference feature amount 41. However, a different feature amount may be computed as each of the image feature amount and the reference feature amount 41.
In a vehicle determination process in step S106, the vehicle determination unit 23 compares the reference feature amount 41 acquired in step S104 with the image feature amount computed in step S105 and determines whether the image feature amount and the reference feature amount 41 are similar.
Specifically, the vehicle determination unit 23 reads the reference feature amount 41 from the storage 122 in step S104 and reads, from the memory 121, the image feature amount computed in step S105. In the first embodiment, the vehicle determination unit 23 compares the image feature amount that has been read with the reference feature amount 41 that has been read, using an SVM, and a similarity degree is computed. Then, the vehicle determination unit 23 determines whether or not the similarity degree is higher than a threshold value.
If the image feature amount and the reference feature amount have been determined to be similar, the vehicle determination unit 23 causes the procedure to proceed to step S107. On the other hand, if the image feature amount and the reference feature amount have been determined not to be similar, the vehicle determination unit 23 causes the procedure to proceed to step S108.
Herein, the SVM has been employed for determining the similarity of the feature amounts. A different method, however, may be employed for determining the similarity of the feature amounts.
That is, as illustrated in
In a vehicle determination process in step S107, the vehicle determination unit 23 determines that the surrounding vehicle 200 is photographed in the partial region 61 selected in step S102. That is, the vehicle determination unit 23 determines that the surrounding vehicle 200 is present in the partial region 61. Then, the vehicle determination unit 23 identifies the position of the surrounding vehicle 200 based on the position of the partial region 61 in the image information 42.
The vehicle determination unit 23 transmits identifying information 44 indicating the identified position of the surrounding vehicle 200 to the vehicle control unit 32. Then, the vehicle determination unit 23 returns the procedure to step S102.
The vehicle control unit 32 controls traveling of the vehicle 100, using the identifying information 44 that has been transmitted. If the surrounding vehicle 200 has been detected near the traveling direction of the vehicle 100, for example, the vehicle control unit 32 performs brake control of the vehicle 100, thereby avoiding collision with the surrounding vehicle 200.
In a non-vehicle determination process in step S108, the vehicle determination unit 23 determines that no vehicle is photographed in the partial region 61 selected in step S102. That is, the vehicle determination unit 23 determines that no surrounding vehicle 200 is present in the partial region 61. Then, the vehicle determination unit 23 returns the procedure to step S102.
If the direction identification unit 21 has selected all the regions indicated by the image information 42 as the partial regions 61 in step S102, the procedure is finished.
As mentioned above, the vehicle determination apparatus 10 according to the first embodiment determines whether or not the surrounding vehicle 200 is photographed in the partial region 61, using the reference feature amount 41 corresponding to the traveling direction of the surrounding vehicle 200 in the partial region 61. This can detect the surrounding vehicle 200 from the image information 42 at high speed and with high accuracy.
In the first embodiment, the function of each unit of the vehicle determination apparatus 10 has been implemented by the software. As a first variation example, however, the function of each unit of the vehicle determination apparatus 10 may be implemented by hardware. A difference of this first variation example from the first embodiment will be described.
A configuration of a vehicle determination apparatus 10 according to the first variation example will be described with reference to
When the function of each unit is implemented by the hardware, the vehicle determination apparatus 10 includes a processing circuit 14 in place of the processor 11 and the storage device 12. The processing circuit 14 is an electronic circuit dedicated to implementing the function of each unit in the vehicle determination apparatus 10 and a function of the storage device 12.
The processing circuit 14 is assumed to be a single circuit, a composite circuit, a programmed processor, a parallel-programmed processor, a logic IC, a GA (Gate Array), an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate Array).
The function of each unit may be implemented by one processing circuit 14, or the function of each unit may be distributed into a plurality of the processing circuits 14, for implementation.
A part of the functions may be implemented by hardware, and the other functions may be implemented by software, as a second variation example. That is, the part of the functions of the respective units in the vehicle determination apparatus 10 may be implemented by the hardware and the other functions may be implemented by the software.
The processor 11, the storage device 12, and the processing circuit 14 are collectively referred to as “processing circuitry”. That is, the functions of the respective units are implemented by the processing circuitry.
A second embodiment is different from the first embodiment in that a region of a parallel lane 51 and a region of an opposite lane 52 are identified when there are a plurality of lanes on each side of a road. In the second embodiment, this difference will be described.
In the second embodiment, a description will be given about a case where there are two lanes on each side of the road. Even a case where there are three or more lanes on each side of the road, however, can be handled, based on a similar concept.
A configuration of a vehicle determination apparatus 10 according to the second embodiment will be described with reference to
The vehicle determination apparatus 10 includes, in addition to the hardware configuration of the vehicle determination apparatus 10 illustrated in
The communication interface 15 is a device including a receiver to receive data and a transmitter to transmit data. As a specific example, the communication interface 15 is a communication chip or a NIC (Network Interface Card).
The positioning sensor 16 is a device to receive a positioning signal that has been transmitted by a positioning satellite such as a GPS satellite and can identify the position of a vehicle 100. A quasi-zenith satellite to take an orbit where the quasi-zenith satellite stays in the sky over a specific region for a long period of time may be used as the positioning satellite. Use of the quasi-zenith satellite allows high-precision position information to be computed.
The vehicle determination apparatus 10 includes, in addition to the functional components of the vehicle determination apparatus 10 illustrated in
Operations of the vehicle determination apparatus 10 according to the second embodiment will be described with reference to
The operations of the vehicle determination apparatus 10 according to the second embodiment correspond to a vehicle determination method according to the second embodiment. The operations of the vehicle determination apparatus 10 according to the second embodiment correspond to a vehicle determination program procedure according to the second embodiment.
The operations of the vehicle determination apparatus 10 according to the second embodiment will be outlined with reference to
In each of
As illustrated in
The operations of the vehicle determination apparatus 10 according to the second embodiment will be described in detail, with reference to
In the second embodiment, processes illustrated in
In a photographed image acquisition process in step S201, the direction identification unit 21 acquires image information 42, as in step S101 in
In a position detection process in step S202, the lane identification unit 24 acquires a positioning signal via the positioning sensor 16, thereby identifying the position of a vehicle 100. Identification of the position of the vehicle 100 corresponds to identification of the position of a camera 31.
Specifically, the lane identification unit 24 receives the positioning signal from a positioning satellite such as a GPS via the positioning sensor 16 and receives, via the positioning sensor 16, positioning correction information for the positioning satellite from a quasi-zenith satellite that takes an orbit in which the quasi-zenith satellite stays in the sky over a specific region. The lane identification unit 24 corrects a position that is computed by the received positioning signal, using the positioning correction information and identifies the position. The lane identification unit 24 writes the identified position into the memory 121.
The positioning correction information is constituted from observation data at an electronic reference point set on the ground, satellite orbit information, a tropospheric delay model, and an ionospheric delay model. The tropospheric delay model indicates an effect that the speed of the radio wave of the positioning signal is reduced more in a troposphere close to the ground than in vacuum due to passage of the positioning signal through the atmosphere and an effect that the distance of a propagation path increases more than that of a direct line due to slight bending of the propagation path. The ionospheric delay model indicates an effect that the electron density of an ionosphere in the vicinity of the altitude of 250 to 400 km from the ground changes due to solar radiation, and the speed of the radio wave of the GPS positioning signal changes, depending on the electron density. Use of the positioning correction information can identify the position of the vehicle 100 with a precision of units of several centimeters if the road has no upper obstruction such as a tunnel.
Herein, the position of the photographing has been identified, using the positioning signal. However, it may be so arranged that a white line position on the road is detected, using a sensor such as a LIDAR (Light Detection and Ranging, Laser Imaging Detection and Ranging) sensor, and the position of the vehicle 100 is identified based on the white line position.
In a map acquisition process in step S203, the lane identification unit 24 acquires map information around the position of the vehicle 100 identified in step S202.
Specifically, the lane identification unit 24 reads the position identified in step S202 from the memory 121. Then, the lane identification unit 24 receives, from an external server, the map information around the position identified in step S202, via the communication interface 15. The map information is information having a precision with which the position of each lane in the road can be identified. Three-dimensional shape information of the road may be included in the map information. The lane identification unit 24 writes the received map information into the memory 121.
Herein, the lane identification unit 24 has received the map information from the external server. However, the map information may be stored in the storage 122, and the lane identification unit 24 may read the map information from the storage 122.
In a travel lane identification process in step S204, the lane identification unit 24 identifies a travel lane in which the vehicle 100 travels.
A specific description will be given with reference to
Assume, for example, that the map information is a numerical expression indicating each line that delimits the lanes, as illustrated in
In a region selection process in step S205, the direction identification unit 21 selects a partial region 61 of a region indicated by the image information 42 acquired in step S201, as in step S102 in
In a direction identification process in step S206, the direction identification unit 21 identifies a traveling direction in which a surrounding vehicle 200 travels in the partial region 61 selected in step S205.
Specifically, the direction identification unit 21 reads, from the memory 121, the partial region 61 selected in step S205 and the travel lane information 45 indicating the travel lane identified in step S204. The direction identification unit 21 determines whether the partial region 61 is a region of a parallel lane 51, a region of an opposite lane 52, or a region of neither the parallel lane 51 nor the opposite lane 52, in consideration of the travel lane indicated by the travel lane information 45 that has been read.
If the partial region 61 has been determined to be the region of the parallel lane 51 or the opposite lane 52, the direction identification unit 21 causes the procedure to proceed to step S207. If the partial region 61 has been determined to be region of neither the parallel lane 51 nor the opposite lane 52, the direction identification unit 21 returns the procedure to step S205 and selects another partial region 61.
A method of determining whether the partial region 61 is the region of the parallel lane 51, the region of the opposite lane 52, or the region of neither the parallel lane 51 nor the opposite lane 52 with reference to
When the vehicle 100 travels in the lane adjacent to a road shoulder 53, a photographed region of a road 50 is shifted to the right by an amount corresponding to one lane, as illustrated in
Then, the direction identification unit 21 sets the region of the parallel lane 51 and the region of the opposite lane 52 for each travel lane of the vehicle 100. When the vehicle 100 travels in the lane adjacent to the opposite lane 52, for example, the direction identification unit 21 sets, as the parallel lane 51, a region from the left end to the center position or a position on a slightly right side of the center position of the image information 42, and sets, as the opposite lane 52, a region from the right end to the center position or a position on a slightly left side of the center position of the image information 42. When the vehicle 100 travels in the lane adjacent to the road shoulder 53, the direction identification unit 21 sets, as the parallel lane 51, a region from a position moved to the right from the left end of the image information 42 by one lane to a position moved to the right from the center position by one lane or a position moved to the right from the slightly right side of the center potion by one lane, and sets, as the opposite lane 52, a range on the right side of the parallel lane 51.
The direction identification unit 21 determines whether the partial region 61 is the region of the parallel lane 51, the region of the opposite lane 52, or the region not included in neither the parallel lane 51 nor the opposite lane 52, according to whether the partial region 61 includes a region of the parallel lane 51, includes a region of the opposite lane 52, or includes neither the region of the parallel lane 51 nor the region of the opposite lane 52.
The processes from step S207 to step S211 are the same as the processes from step S104 to step S108 illustrated in
That is, only when the partial region 61 includes the region of the parallel lane 51 or the region of the opposite lane 52, as illustrated in
As mentioned above, the vehicle determination apparatus 10 according to the second embodiment identifies the travel lane of the vehicle 100, thereby identifying the parallel lane 51 and the opposite lane 52 with good accuracy.
With this arrangement, the processes after step S207 are omitted for a region such as the road shoulder 53 where no surrounding vehicle 200 travels. Therefore, the surrounding vehicle 200 can be detected from the image information 42 at high speed. Further, similarity determination of the feature amount is made, using an appropriate reference feature amount 41. Therefore, the surrounding vehicle 200 can be detected from the image information 42 with good accuracy.
In the second embodiment, the travel lane of the vehicle 100 has been identified in step S204, and the region of the parallel lane 51 and the region of the opposite lane 52 have been identified in step S206, in consideration of the travel lane. As a third variation example, the region of a parallel lane 51 and the region of an opposite lane 52 in a region indicated by image information 42 may be identified, using the position of a vehicle 100 and map information.
A configuration of a vehicle determination apparatus 10 according to the third variation example will be described, with reference to
The vehicle determination apparatus 10 is different from the vehicle determination apparatus 10 illustrated in
Operations of the vehicle determination apparatus 10 according to the third variation example will be described with reference to
In the third variation example, processes illustrated in
The processes from step S301 to step S303 are the same as the processes from step S201 to step S203 in
The process in step S304 is the same as the process in step S205 in
In a direction identification process in step S305, the direction identification unit 21 identifies a traveling direction in which a surrounding vehicle 200 travels in a partial region 61 from the position of the vehicle 100 identified in step S302 and the map information acquired in step S303.
A specific description will be given with reference to
The direction identification unit 21 identifies a focus position (X0, Y0, Z0) of the camera 31 when photographing is performed. The focus position (X0, Y0, Z0) of the camera 31 is identified from the position of the vehicle 100 and a setting parameter of the camera 31 such as a focal length for each pixel.
The direction identification unit 21 acquires, from the map information, positions (X1, Y1, Z1) of some boundary portions of a parallel lane 51 and positions (X2, Y2, Z2) of some boundary portions of an opposite lane 52, and then computes relative positions with respect to the focus position (X0, Y0, Z0) of the camera 31. If the focus position of the camera 31 is an origin (0, 0, 0), a relative position of the parallel lane 51 becomes (X1, Y1, Z1), and a relative position of the opposite lane 52 becomes (X2, Y2, Z2). The direction identification unit 21 computes a pixel position (u1, v1) obtained by projecting the relative position (X1, Y1, Z1) of the parallel lane 51 on the imaging plane of the camera 31 by a perspective projective transformation process. Similarly, the direction identification unit 21 computes a pixel position (u2, v2) obtained by projecting the relative position (X2, Y2, Z2) of the opposite lane 52 on the imaging plane of the camera 31 by the perspective projective transformation process.
The direction identification unit 21 computes each pixel position (u1, v1) with respect to the positions of the boundary positions (X1, Y1, Z1) and computes each pixel position (u2, v2) with respect to the positions of the boundary positions (X2, Y2, Z2). Then, the direction identification unit 21 sets a range to be identified from the respective positions (u1, v1) as the region of the parallel lane 51, and sets a range to be identified from the respective positions (u2, v2) as the region of the opposite lane 52. As a specific example, the direction identification unit 21 connects the positions (u1, v1) that are adjacent to each other, and sets a range enclosed by the adjacent positions (u1, v1) to the region of the parallel lane 51, and connects the positions (u2, v2) that are adjacent to each other, and sets a range enclosed by the adjacent positions (u2, v2) to the region of the opposite lane 52.
Herein, an equation for perspective projective transformation is as given by Expression 1. fx and fy indicate a focal length for each pixel, which is the setting parameter of the camera 31. The pixel position (u1, v1) indicates a pixel position when the center of the imaging plane of the camera device (center position of a photographed image) is set to the position of an origin (0, 0).
With this arrangement, the region of the parallel lane 51 and the region of the opposite lane 52 can be more accurately identified. As a result, the surrounding vehicle 200 can be detected from the image information 42 at high speed and with good accuracy.
A third embodiment is different from the first and second embodiments in that similarity determination of a feature amount is made, using a reference feature amount 41 corresponding to the position of an associated lane being a lane of a partial region 61.
A configuration of a vehicle determination apparatus 10 according to the third embodiment will be described, with reference to
The vehicle determination apparatus 10 is different from the vehicle determination apparatus 10 illustrated in
Operations of the vehicle determination apparatus 10 according to the third embodiment will be described with reference to
The operations of the vehicle determination apparatus 10 according to the third embodiment correspond to a vehicle determination method according to the third embodiment. The operations of the vehicle determination apparatus 10 according to the third embodiment correspond to a vehicle determination program procedure according to the third embodiment.
The operations of the vehicle determination apparatus 10 according to the third embodiment will be outlined with reference to
Referring to
In this case, a right oblique rear portion of a surrounding vehicle 200A that travels in the lane X1 is photographed. The rear of a surrounding vehicle 200B that travels in the lane X2 is photographed. A right oblique front portion of a surrounding vehicle 200C that travels in the lane X3 is photographed. A right oblique front portion of a surrounding vehicle 200D that travels in the lane X4, which is closer to a side portion of the surrounding vehicle 200D than in the case of the surrounding vehicle 200C that travels in the lane X3, is photographed.
That is, according to the relative travel position of each surrounding vehicle 200 with respect to the vehicle 100, a photographed angle differs. Then, the vehicle determination apparatus 10 stores a reference feature amount 41, for each relative position of an associated lane with respect to the travel lane of the vehicle 100 and associated with each partial region 61. Then, using the reference feature amount 41 corresponding to the relative position of the associated lane with respect to the travel lane of the vehicle 100 and associated with the partial region 61, the vehicle determination apparatus 10 determines whether or not the surrounding vehicle 200 is present in the partial region 61 to be processed.
The reference feature amount 41 according to the third embodiment will be described, with reference to
The reference feature amount 41 is computed using reference images corresponding to a target photographing direction and the relative position of an associated lane with respect to the travel lane. In
The reference feature amount 41 corresponding to each class is stored in the storage 122, being associated with each class as illustrated in
In the case of driving to the left of a road, it does not happen in principle that the surrounding vehicle 200 that travels in the opposite lane 52 travels in the same lane and the left-adjacent lane. Therefore, the reference feature amounts 41 for the front of the vehicle corresponding to the same lane and the left-adjacent lane do not need to be stored. Further, in the case of two lanes on each side of a road, it does not happen in principle that the surrounding vehicle 200 that travels in the parallel lane 51 travels in the second-right adjacent lane. Therefore, the reference feature amount 41 for the rear of the vehicle corresponding to the second right-adjacent lane does not need to be stored.
The operations of the vehicle determination apparatus 10 according to the third embodiment will be described in detail, with reference to
In the third embodiment, processes illustrated in
The processes from step S401 to step S406 are the same as the processes from step S201 to step S206 illustrated in
In an associated lane identification process in step S407, the lane identification unit 24 identifies an associated lane associated with a partial region 61 selected in step S405.
Specifically, the lane identification unit 24 identifies a range of each lane on the imaging place of a camera 31 using the method described with reference to
If the partial region 61 includes regions of two lanes, both of the two lanes are identified as associated lanes.
In a feature amount acquisition process in step S408, the feature acquisition unit 22 acquires a reference feature amount 41 corresponding to the traveling direction of a surrounding vehicle 200 identified in step S406 and the associated lane identified in step S407. That is, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the traveling direction of the surrounding vehicle 200 and the relative position of the associated lane with respect to a travel lane. The feature acquisition unit 22 writes, into the memory 121, the reference feature amount 41 that has been read.
As a specific example, when the traveling direction of the surrounding vehicle 200 is the same as that of a vehicle 100 and the associated lane is the left-adjacent lane of the travel lane, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the same direction as that of the vehicle 100 and the left-adjacent lane. That is, the feature acquisition unit 22 reads a reference feature amount 41C in
When the partial region 61 includes regions of two lanes, the feature acquisition unit 22 acquires the reference feature amounts 41 corresponding to both of two associated lanes.
As mentioned above, the vehicle determination apparatus 10 according to the third embodiment determines whether or not the surrounding vehicle 200 is photographed in the partial region 61, using the reference feature amount 41 corresponding to the traveling direction of the surrounding vehicle 200 in the partial region 61 and the relative position of the associated lane with respect to the travel lane. With this arrangement, the surrounding vehicle 200 can be detected from image information 42 at high speed and with high accuracy.
In the third embodiment, classification of the reference feature amount 41 has been performed, according to the relative position of the associated lane with respect to the travel lane. As a fourth variation example, classification of a reference feature amount 41 may be performed according to a distance between a vehicle 100 and each surrounding vehicle 200 in a traveling direction as well as the relative position of an associated lane with respect to a travel lane.
To take an example, even if the surrounding vehicle 200 is the one that travels in a left-adjacent lane, an angle at which the surrounding vehicle 200 is photographed differs between when the surrounding vehicle 200 is positioned separated from the vehicle 100 and when the surrounding vehicle 200 is positioned close to the vehicle 100.
Then, the vehicle determination apparatus 10 classifies each reference feature amount 41 according to the distance between the vehicle 100 and the surrounding vehicle 200 in the traveling direction as well the relative position of the associated lane with respect to the travel lane and stores the classified reference feature amount 41 in the storage 122. Then, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the distance between the vehicle 100 and the surrounding vehicle 200 in the traveling direction.
Herein, the surrounding vehicle 200 positioned separated from the vehicle 100 is photographed in an upper region of image information 42, and the surrounding vehicle 200 positioned close to the vehicle 100 is photographed in a lower region of the image information 42. Accordingly, by acquiring the reference feature amount 41 corresponding to the position of a partial region 61 in the an upper or a lower direction of a region indicated by the image information 42 by the feature acquisition unit 22, the reference feature amount 41 corresponding to the distance between the vehicle 100 and the surrounding vehicle 200 in the traveling direction can be acquired.
With this arrangement, similarity determination of a feature amount is made using a more appropriate reference feature amount 41. Therefore, the surrounding vehicle 200 can be therefore detected from the image information 42 with high accuracy.
In the third embodiment, it has been considered whether the surrounding vehicle 200 travels in the same direction as the vehicle 100 or whether the surrounding vehicle 200 travels in the opposite direction to the vehicle 100. A description will be directed to a case where there is an intersection in front of a vehicle 100, as a fifth variation example.
As illustrated in
Then, the vehicle 100 stores a reference feature amount 41 for the intersection in the storage 122. The reference feature amount 41 for the intersection is computed by using reference images obtained by photographing from right and left side portions of the vehicles and reference images obtained by photographing from right and left oblique front and rear portions of the vehicles. Then, in step S407 in
With this arrangement, the surrounding vehicle 200 at the intersection can be detected at high speed and with high accuracy.
A fourth embodiment is different from the first to third embodiments in that similarity determination of a feature amount is made, using a reference feature amount 41 for a surrounding vehicle 200 that has been photographed with a part thereof hidden.
In the fourth embodiment, a description will be given about a case where a function is added to the third embodiment. The function may be, however, added to the first embodiment or the second embodiment.
A configuration of a vehicle determination apparatus 10 according to the fourth embodiment will be described with reference to
The vehicle determination apparatus 10 includes, in addition to the functional components of the vehicle determination apparatus 10 illustrated in
Operations of the vehicle determination apparatus 10 according to the fourth embodiment will be described, with reference to
The operations of the vehicle determination apparatus 10 according to the fourth embodiment correspond to a vehicle determination method according to the fourth embodiment. The operations of the vehicle determination apparatus 10 according to the fourth embodiment correspond to a vehicle determination program procedure according to the fourth embodiment.
The operations of the vehicle determination apparatus 10 according to the fourth embodiment will be outlined with reference to
Referring to
As illustrated in
The reference feature amount 41 according to the fourth embodiment will be described with reference to
The reference feature amount 41 is computed, using reference images corresponding to a target photographing direction, the relative position of an associated lane with respect to a travel lane, and presence or absence of the shielded region 47. Referring to
The reference feature amount 41 corresponding to each class is stored in the storage 122, being associated with each class as illustrated in
The operations of the vehicle determination apparatus 10 according to the fourth embodiment will be described in detail with reference to
In the fourth embodiment, processes illustrated in
The processes from step S501 to step S507 are the same as the processes from step S401 to step S407 illustrated in
In a shielding determination process in step S508, the shielding determination unit 25 determines whether or not a partial region 61 selected in step S505 includes a shielded region 47.
A specific description will be given with reference to
In a feature amount acquisition process in step S509, the feature acquisition unit 22 acquires a reference feature amount 41 corresponding to the traveling direction of a surrounding vehicle 200 that has been identified in step S506, an associated lane that has been identified in step S507, and presence or absence of the shielded region 47 determined in step S508.
As a specific example, if the traveling direction of the surrounding vehicle 200 is the same direction as that a vehicle 100, the associated lane is the left-adjacent lane of a travel lane, and the shielded region 47 is present, the feature acquisition unit 22 acquires the reference feature amount 41 corresponding to the same direction as that of the vehicle 100, the left-adjacent line, and the presence of the shielded region 47. That is, the feature acquisition unit 22 reads, from the storage 122, a reference feature amount 41C2 in
As mentioned above, the vehicle determination apparatus 10 according to the fourth embodiment determines whether or not the surrounding vehicle 200 is photographed in the partial region 61, using the reference feature amount 41 corresponding to whether or not the partial region 61 includes the shielded region 47. With this arrangement, the surrounding vehicle 200 can be detected from the image information 42 at high speed and with high accuracy.
In each of the second to fourth embodiments, the function of each unit of the vehicle determination apparatus 10 has been implemented by the software, as in the first embodiment. The function of each unit of the vehicle determination apparatus 10, may be, however, implemented by the hardware, as in the first variation example. Further, as in the second variation example, a part of the functions of the vehicle determination apparatus 10 may be implemented by the hardware, and the other functions may be implemented by the software.
The above description has been given about the embodiments of the present invention. Some of these embodiments and the variation examples may be carried out in combination. Alternatively, any one or some of the embodiments and the variation examples may be partially carried out. The present invention is not limited to the embodiments and the variation examples described above, and various modifications are possible as necessary.
10: vehicle determination apparatus; 11: processor; 12: storage device; 13: camera interface; 14: processing circuit; 15: communication interface; 16: positioning sensor; 21: direction identification unit; 22: feature acquisition unit; 23: vehicle determination unit; 24: lane identification unit; 25: shielding determination unit; 31: camera; 32: vehicle control unit; 321: vehicle control interface; 322: sensor ECU; 323: vehicle ECU; 41: reference feature amount; 42: image information; 43: direction information; 44: identifying information; 45: travel lane information; 46: associated lane information; 47: shielded region; 50: road; 51: parallel lane; 52: opposite lane; 53: road shoulder; 61: partial region; 62: cell; 100: vehicle; 200: surrounding vehicle
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2016/052109 | 1/26/2016 | WO | 00 |