Patent Application
20090123031
The present invention generally relates to a system and method of awareness detection, and more particularly, to a system and method of detecting awareness based upon a subject's eye movement.
Video imaging systems have been proposed for use in vehicles to monitor a subject person such as the driver and other passengers in the vehicle. Some proposed video imaging systems include one or two cameras focused on the driver of the vehicle to capture images of the driver's face. The video images are processed generally using computer vision and pattern recognition techniques to determine various facial characteristics of the driver including position, orientation, and movement of the driver's eyes, face, and head. Some advanced eye monitoring systems process the captured images to determine eye closure, such as open, half-open (half-closed), and closed states of the eye(s).
By knowing the driver's facial characteristics, vehicle control systems can provide enhanced vehicle functions. For example, a vehicle control system can monitor one or both eyes of the subject driver and determine a condition in which the driver appears to be fatigued or drowsy based on statistical analysis of the cumulated results of open or closed state of the eye(s) over time. Generally, standard human factor measures such as PerClos (percentage of eye closure) and AveClos (average of eye closure) could be used to determine the drowsiness state of the driver. For instance, if the AveClos value is determined to be above a certain threshold, the system may initiate countermeasure action(s) to alert the driver of the driver drowsy condition and/or attempt to awaken the driver.
Some proposed vision-based imaging systems that monitor the eye(s) of the driver of a vehicle require infrared (IR) illumination along with visible light filters to control scene brightness levels inside of the vehicle cockpit. One such driver monitoring system produces bright and dark eye conditions that are captured as video images, which are processed to determine whether the eye is in the open position or closed position. Such prior known driver eye monitoring systems generally require specific setup of infrared illuminators on and off the optical camera axis. In addition, these systems are generally expensive, their setup in a vehicle is not practical, and they may be ineffective when used in variable lighting conditions, especially in bright sunny conditions. Further, variations in eyelash contrast and eye iris darkness levels for different subject persons may cause such prior systems to make erroneous eye state discrimination decisions.
According to one aspect of the present invention, an awareness detection system is provided. The system includes an imaging device positioned to obtain a plurality of images of at least a portion of a subject's head, and an awareness processor in communication with the imaging device. The awareness processor receives the plurality of images from the imaging device and performs the steps including classifying at least one image of the plurality of images based upon a head pose of at least a portion of the subject's head with respect to at least one image, monitoring movement of at least one eye of the subject if the at least one image is classified as a predetermined classification, and determining an awareness state of the subject based upon the monitored movement of the at least one eye, wherein the movement of at least one eye of the subject is monitored over at least two images obtained by the imaging device.
According to another aspect of the present invention, a method of detecting awareness of a subject is provided. The method includes the steps of obtaining a plurality of images of at least a portion of a subject, classifying at least one image of the plurality of images based upon a head pose of at least a portion of the subject's head with respect to at least one image, wherein the classification of the image includes one of frontal and non-frontal, monitoring movement of at least one eye of the subject if the at least one image is classified as a predetermined classification, and determining an awareness state of the subject based upon the monitored movement of at least one said eye, such that said awareness state includes one of distracted and non-distracted, wherein the movement of the at least one eye is monitored over at least two images.
According to yet another aspect of the present invention, a method of detecting awareness of a subject is provided. The method includes the steps of obtaining a plurality of images of at least a portion of a subject, wherein the subject is an occupant in a vehicle, classifying at least one image of the plurality of images based upon a head pose of at least a portion of the subject's head with respect to at least one image, wherein the subject in the at least one image is classified as one of frontal and non-frontal, monitoring movement of at least one eye of the subject if the at least one image is classified as the frontal classification, and determining an awareness state of the subject based upon said the monitored movement of the at least one eye, such that said awareness state is one of distracted and non-distracted, wherein the movement of the at least one eye is monitored over at least two images.
These and other features, advantages and objects of the present invention will be further understood and appreciated by those skilled in the art by reference to the following specification, claims and appended drawings.
The present invention will now be described, by way of example, with reference to the accompanying drawings, in which:
Referring to
The imaging device 16 is shown located generally in front of a driver's seat 20 in the front region of the passenger compartment 14. According to one embodiment, the imaging device 16 is a non-intrusive system that is mounted in the instrument cluster. However, the imaging device 16 may be mounted in other suitable locations onboard the vehicle 12, which allow for acquisition of images capturing the subject's 18 head. By way of explanation and not limitation, the imaging device 16 may be mounted in a steering assembly 22 or mounted in a dashboard 24. While a single imaging device 16 is shown and described herein, it should be appreciated by those skilled in the art that two or more imaging devices may be employed in the awareness detection system 10.
The imaging device 16 can be arranged so as to capture successive video image frames of the region where the subject 18, which in a disclosed embodiment is typically driving the vehicle 12, is expected to be located during normal vehicle driving. More particularly, the acquired images capture at least a portion of the subject's 18 face, which can include one or both eyes. The acquired images are then processed to determine characteristics of the subject's 18 head, and to determine the awareness of the subject 18. For purposes of explanation and not limitation, the detected awareness of the subject 18 can be used to control other components of the vehicle 12, such as, but not limited to, deactivating a cruise control system, activating an audio alarm, the like, or a combination thereof.
According to one embodiment, the awareness detection system 10 can include a light illuminator 26 located forward of the subject 18, such as in the dashboard 24, for illuminating the face of the subject 18. The light illuminator 26 may include one or more infrared (IR) light emitting diodes (LEDs). Either on-access or off-access LEDs may be employed (e.g., no specific IR setup is required, in particular). The light illuminator 26 may be located anywhere onboard the vehicle 12 sufficient to supply any necessary light illumination to enable the imaging device 16 to acquire images of the subject's 18 head.
With regards to
The awareness processor 28 can include a frame grabber 32 for receiving the video output frames generated by the imaging device 16. The awareness processor 28 can also include a digital signal processor (DSP) 34 for processing the acquired images. The DSP 34 may be a floating point or fixed point processor. Additionally, the awareness processor 28 can include memory 36, such as random access memory (RAM), read-only memory (ROM), and other suitable memory devices, as should be readily apparent to those skilled in the art. The awareness processor 28 is configured to perform one or more awareness detection routines for controlling activation of the light illuminator 26, controlling the imaging device 16, processing the acquired images to determine the awareness of the subject 18, and applying the processed information to vehicle control systems, such as the host processor 30.
The awareness processor 28 may provide imager control functions using a control RS-232 logic 38, which allows for control of the imaging device 16 via camera control signals. Control of the imaging device 16 may include automatic adjustment of the orientation of the imaging device 16. For purposes of explanation and not limitation, the imaging device 16 may be repositioned to focus on an identifiable feature, and may scan a region in search of an identifiable feature, including the subject's 18 head, and more particularly, one of the eyes of the subject 18. Also, the imager control may include adjustment of the focus and magnification as may be necessary to track identifiable features of the subject 18.
According to one embodiment, the awareness processor 28 is in communication with the imaging device 16, such that the awareness processor 28 receives the plurality of images from the imaging device 16, and performs the step of classifying at least one image of the plurality of images based upon at least a portion of the subject's 18 head. Additionally, the awareness processor 28 performs the steps of monitoring movement of at least one eye of the subject 18 if the at least one image is classified as a predetermined classification, and determining an awareness state of the subject 18 based upon the monitored movement of the at least one eye, as described in greater detail herein.
According to one embodiment, the subject 18 in at least one of the images is classified as frontal or non-frontal. For purposes of explanation and not limitation, the subject 18 is classified as frontal. If it is determined that the subject's 18 head is between approximately plus/minus twenty degrees (20°) from a straight-forward position, and the image is classified as non-frontal if the subject's 18 head is determined to be outside the plus/minus twenty degrees (20°) range. According to a disclosed embodiment, the predetermined classification for monitoring the movement of at least one eye of the subject 18 is a frontal classification, such that the eyes of the subject 18 are monitored only if the image is classified with a frontal classification.
In reference to
However, if it is determined at decision step 108 that the image is classified as frontal, the method 100 proceeds to step 112, wherein at least one eye of the subject 18 is located and is monitored over a plurality of images. Thus, the movement of the eyes can be monitored over a period of time to determine a pattern of eye movement. At step 114, the subject is classified based upon the detected eye movement at step 112. At decision step 116, it is determined if the subject 18 is distracted based upon the classification of the subject's 18 eye movements. If it is determined at decision step 116 that the subject 18 is distracted, then the method 100 can proceed to step 110, wherein counter measures are activated, according to one embodiment. According to an alternate embodiment, if it is determined at decision step 116 that the subject 18 is distracted, the method can then end at step 111. However, if it is determined that the subject 18 is not distracted at decision step 116, then the method 100 can return to step 104 to obtain an image. It should be appreciated by those skilled in the art that if it is determined at decision step 116 that the subject 18 is distracted, prior to or after counter measures are activated at step 110, the method 100 can return to step 104 to obtain an image.
According to one embodiment, the image classification at step 106 can include a head pose estimation, such as, but not limited to, an appearance based head pose estimation or a geometric based head pose estimation. According to a disclosed embodiment, predetermined facial features of the subject 18 can be extracted, such as, but not limited to, eyes, nose, the like, or a combination thereof. Thus, a face box or portion of the image to be analyzed or monitored can be determined based upon the extracted features. According to one embodiment, facial features of the subject 18 can be extracted from the image and compared to a three-dimensional (3D) head pose model. Alternatively, a head pose estimation can include detecting a face of the subject 18 and classifying the detected face online using a classification rule (i.e., distance) that can employ head pose appearance models built or constructed off-line. The head pose appearance models can be built during a testing phase, a development phase, or an experimental phase, according to one embodiment.
In reference to
At step 126, the three ROIs defined in step 124 are extracted from the image, and resized to a predetermined size, such that all three head boxes are the same size. At step 128, the awareness processor 28 processes the ROIs. According to a disclosed embodiment, the awareness processor 28 processes the ROIs by applying affine transform and histogram equalization processing to the image. It should be appreciated by those skilled in the art that other suitable image processing techniques can be used additionally or alternatively.
At step 130, each of the ROIs are designated or classified, wherein, according to one embodiment, the ROIs are given two classifications for two models, such that a first model is a normal pose model, and a second model is an outlier model. At step 132, the classifications results for the two models are stored. Typically, the classifications given to each of the head boxes for both the first and second classifications are left, front, or right.
At decision step 134, it is determined if the awareness processor 28 has processed or completed all the ROIs, such that the three ROIs have been classified and the results of the classification have been stored. If it is determined at decision step 134 that all three ROIs have not been completed, then the analysis 106 returns to step 126. However, if it is determined at decision step 134 that the awareness processor 28 has completed the three ROIs, then the analysis 106 proceeds to step 136. At step 136, the classifications are compared. According to a disclosed embodiment, the three ROIs each have two classifications, which are either left, front, or right, and thus, the number of front, left, and right votes can be determined. By way of explanation and not limitation, each ROI is classified as left, right, or front for both the normal pose model and the outlier model, and thus, there are a total of six classifications for the three ROIs, according to this embodiment. According to an alternate embodiment, each captured image has eighteen classifications, such that three ROIs at three different scales are constructed, wherein each ROI has three models and each model has two classifications. At step 138, the classification with the most votes is used to classify the image, and the classification analysis 106 then ends at step 140.
For purposes of explanation and not limitation, the outlier model can include a frontal image of the subject 18, such that the frontal classification is determined by patterns in the image that are not the subject's 18 eyes. The patterns can be, but are not limited to, the subject's 18 head, face, and neck outline with respect to the headrest. Thus, a head pose classification or analysis can be performed using such patterns or the like.
According to one embodiment, the awareness processor 28 determines the awareness state of the subject 18 as being one of distracted or non-distracted. According to a disclosed embodiment, the eyes of the subject 18 are monitored and plotted in a Cartesian coordinate plane in order to monitor the movement of the eye. Thus, a pattern of eye movement can be detected, such that a subject 18 can be determined to be distracted. According to one embodiment, at least one of the center of the subject's 18 eye is detected and tracked on an X-axis and Y-axis of a Cartesian coordinate plane, the movement of the subject's 18 eyelid (i.e., tightening or widening), an iris or pupil of the subject's 18 eye, or a combination thereof. Thus, the subject's 18 eye can be tracked as a whole, even though only a portion of the eye is detected or tracked. It should be appreciated by those skilled in the art that the detected eye movement can be tracked to classify the subject 18, even though the subject 18 is in a frontal position.
Exemplary systems and methods of monitoring the eye movement are U.S. patent application Ser. No. 11/452,871 (DP-315413), entitled “METHOD OF TRACKING A HUMAN EYE IN A VIDEO IMAGE,” which is hereby incorporated herein by reference, and U.S. patent application Ser. No. 11/452,116 (DP-313993), entitled “IMPROVED DYNAMIC EYE TRACKING SYSTEM,” which is hereby incorporated herein by reference. An exemplary system and method of classifying an image of a subject as frontal and non-frontal is U.S. patent application Ser. No. 11/890,066 (DP-315567), entitled “SYSTEM AND METHOD OF AWARENESS DETECTION,” which is hereby incorporated herein by reference.
With regards to
If it is determined at decision step 156 that the eye motion of the subject 18 is greater than the threshold value T1, then the motion analysis 112 proceeds to step 158, wherein the counter is reset. According to one embodiment, when the counter is reset, the counter value equals zero. The motion analysis 112 then proceeds to step 152, wherein the subject 18 is classified as distracted, and the motion analysis 112 then ends, and the method 100 can continue, as set forth above.
However, it if is determined at decision step 156 that the eye motion is not above the threshold value T1, then the motion analysis 112 proceeds to step 160, wherein the counter is incremented. According to a disclosed embodiment, the counter is incremented by a value of one (1). At decision step 162, it is determined if the value of the counter is less than a second threshold value T2. If it is determined at decision step 162 that the counter value is less than the second threshold value T2, then the motion analysis 112 proceeds to step 152, wherein the subject 18 is classified as distracted, and the motion analysis 112 then ends, and the method 100 can continue, as set forth above. If it is determined at decision step 162 that the counter value is greater than the second threshold value T2, then the motion analysis 112 proceeds to step 164, wherein the subject 18 is classified as attentive. The motion analysis 112 then ends, and the method 100 can continue, as set forth above.
According to one embodiment, the value of the counter is incremented in order to prevent transient noise, such as, but not limited to, slight eye movement from affecting the classification of the subject 18. Thus, the value of the counter increases the longer the eye of the subject 18 has remained still and/or in substantially the same position. According to a disclosed embodiment, a high value in the counter represents that the eye of the subject 18 has remained substantially still, and thus, the subject 18 is attentive. By contrast, a low value in the counter can represent that the subject's 18 eye is moving, and thus, the subject 18 is distracted, according to one embodiment. Therefore, the threshold values T1,T2 can be predetermined accordingly, according to one embodiment.
According to one embodiment, the monitored eye motion is based upon the length of an X and Y motion vector. Typically, the motion vector is measured in pixels, according to a disclosed embodiment. According to one embodiment, the threshold value T1 is predetermined, but is dependent upon how zoomed in the image is of the subject's 18 head. The threshold value T2 represents 0.5 seconds, such that if the rate of conducting motion analysis 112 is ten times per second, the second threshold value T2 would be five (5), according to one embodiment.
By way of explanation and not limitation, in operation, the awareness detection system 10 and method 100 determine the awareness of a subject 18, who can be a driver of a vehicle 12. The obtained image is first classified as frontal or non-frontal, and thus, if a non-frontal classification is designated, then it can be determined that the subject 18 is distracted. However, if the image is classified as frontal, then the eye movement of the subject 18 is monitored in order to determine if the subject 18 is distracted or non-distracted, since the subject 18 can have a frontal head position while being distracted based upon eye movement without head movement.
Advantageously, the awareness detection system 10 and method 100 accurately determine the awareness of a subject 18 based upon a fusion logic, such that the subject's 18 head positioning and eye movement. By additionally monitoring the eye movement of the subject 18 under predetermined circumstances, the awareness state of the subject 18 can more accurately be determined than if the determination was made solely on the head positioning of the subject 18. It should be appreciated by those skilled in the art that other motion detection techniques and eye detection or tracking techniques can be used in the fusion logic.
The above description is considered that of preferred embodiments only. Modifications of the invention will occur to those skilled in the art and to those who make or use the invention. Therefore, it is understood that the embodiments shown in the drawings and described above are merely for illustrative purposes and not intended to limit the scope of the invention, which is defined by the following claims as interpreted according to the principles of patent law, including the doctrine of equivalents.
