Patent Application
20180286058
This application is based upon and claims the benefit of priority from British Patent Application Number 1705406.5 filed Apr. 4, 2017, the entire contents of which are incorporated by reference.
This disclosure relates to methods and systems for determining the roughness of surfaces.
Optical methods of surface roughness characterisation are well-established. A known approach is to illuminate a surface with coherent light from a laser. The light scattered from the surface may then be directly imaged, allowing capture of a speckle pattern. As the speckle pattern is caused by constructive and destructive interference of the laser light, which is at least in part dependent upon the surface geometry down to the wavelength scale, the image of the speckle pattern may be processed to estimate the surface roughness.
However, problems exist in terms of optimising the efficiency of the image processing pipeline such that the technique may be used for on-line surface roughness measurement.
The invention is directed towards methods of and systems for determining surface roughness.
The method includes obtaining an image of speckle caused by the scattering of coherent light from a surface. Then, the image is converted into a binary image according to a threshold, thereby classifying pixels below the threshold as background pixels and pixels above the threshold as foreground pixels. Then, one or more regions of connected foreground pixels are identified in the binary image, in which any two foreground pixels in a region are joined by a continuous path of foreground pixels. Then, the total number of regions identified is evaluated, and the number of pixels in the largest region is evaluated, each of which correlate with surface roughness. An indication of the surface roughness of the surface is then outputted.
There is also provided a measurement system which implements the method, and instructions executable by a computer which cause the computer to perform the method.
Embodiments will now be described by way of example only with reference to the accompanying drawings, which are purely schematic and not to scale, and in which:
A turbofan engine is shown in
The engine 101 has a principal and rotational axis A-A and comprises, in axial flow series, an air intake 102, a propulsive fan 103, an intermediate pressure compressor 104, a high-pressure compressor 105, combustion equipment 106, a high-pressure turbine 107, an intermediate pressure turbine 108, a low-pressure turbine 109, and an exhaust nozzle 110. A nacelle 111 generally surrounds the engine 101 and defines both the intake 102 and the exhaust nozzle 110.
The engine 101 works in the conventional manner so that air entering the intake 102 is accelerated by the fan 103 to produce two air flows: a first air flow into the intermediate pressure compressor 104 and a second air flow which passes through a bypass duct 112 to provide propulsive thrust. The intermediate pressure compressor 104 compresses the air flow directed into it before delivering that air to the high pressure compressor 105 where further compression takes place.
The compressed air exhausted from the high-pressure compressor 105 is directed into the combustion equipment 106 where it is mixed with fuel and the mixture combusted. The resultant hot combustion products then expand through, and thereby drive the high pressure turbine 107, intermediate pressure turbine 108, and low pressure turbine 109 before being exhausted through the nozzle 110 to provide additional propulsive thrust. The high pressure turbine 107, intermediate pressure turbine 108, and low pressure turbine 109 drive respectively the high pressure compressor 105, intermediate pressure compressor 104, and fan 103, each by a suitable interconnecting shaft.
As the fan 103 is the one of the first components encountered by air entering the intake 102 of the engine 101, it is particularly important to ensure that the blades 201 of the fan 103, identified in
Many developments have therefore taken place in terms of improving fan blade geometry (i.e. on the macro scale). This has resulted in fan blades, such as those forming part of the fan 103, having complex surfaces with high degrees of variation in curvature.
However, it is also important to ensure that air flow is optimised by ensuring the surface roughness of the fan blades 201 is within acceptable limits (i.e. on the micro scale).
A measurement system for determining the roughness of a surface according to an aspect of the present invention is illustrated in
The measurement system is indicated generally at 301, and may be employed for determining the roughness of the fan blade 201 to ensure its surface roughness meets the design specification. However, it will be appreciated that the principles employed by the system allow it to be used for measurement of the roughness of any surface, such as bearings, or the ball of a prosthesis, etc.
Measurement system 301 includes a fixture 302 in which a component may be mounted, such as the fan blade 201 in this example. The exact configuration of the fixture 302 will be understood by those skilled in the art as not being central to the present invention, and instead any fixture in which the component whose roughness is being measured may be held can be used.
A coherent illumination device such as, in this example, a laser 303 is provided to illuminate the surface of the component being analysed. In the present example, the laser 303 is an infrared laser configured to output coherent light at a wavelength of 1064 nanometres at a power of 80 watts. It will be appreciated that different wavelengths and/or different output powers may be adopted depending upon the particular application of the measurement system. Furthermore, the beam width may be selected to match the desired sample area and thus is dependent on the component and the parameters of the testing being undertaken. Thus in an example, an appropriate optical arrangement may be provided to broaden or narrow, and then collimate the light emanating from the laser 303.
In operation, laser light from the laser 303 is directed towards to surface of the component, and is scattered. Due to the generally rough surface of the component, the path length of the wavefronts is such that interference occurs and the intensity of the light varies across the scattered light field. Thus the scattered light field is related to the surface geometry which caused the scattering.
The scattered light field is therefore imaged by an imaging device, which in this example is a camera 304. The camera 304 is in the present example an infrared-sensitive indium gallium arsenide (InGaAs) camera, which in the present example is fitted with a 25 millimetre lens with a fixed aperture of f/16. Of course, it will be appreciated by those skilled in the art that the particular imaging device may be chosen in dependence upon the specifications of the illumination device. In addition, or alternatively, the choice of lens aperture and focal length may be varied, possibly in dependence on the illumination level and/or the size of component being imaged.
The measurement system 301 further comprises a processing device such as, in this example, a personal computer 305 to store and process the images obtained by the camera 304. The devices within the personal computer 305 which enable it to process the images will be described with reference to
In use, the measurement system 301 facilitates determination of the surface roughness by illuminating the surface of the component at a point 306 with coherent light produced by the laser 303. An image of speckle caused by the scattering of the coherent light from the surface is then obtained by the camera 304.
The image is then converted by the personal computer 305 into a binary image according to a threshold, thereby classifying pixels below the threshold as background pixels and pixels above the threshold as foreground pixels.
Connected components are then identified in the binary image. The connected components are regions of connected foreground pixels, in which any two foreground pixels in a region are joined by a continuous path of foreground pixels. The total number of regions identified is then evaluated, along with the number of pixels in the largest region. It has been shown by the present inventors that these values correlate to the actual surface roughness. An indication of the surface roughness of the surface may then he outputted.
Devices within the personal computer 305 are illustrated in
The personal computer 305 comprises a processor such as a central processing unit (CPU) 401. In this instance, central processing unit 401 is a single Intel® Core i7 processor, having four on-die processing cores operating at 3.2 gigahertz. It is of course possible that other processor configurations could be provided, and indeed several such processors could be present to provide a high degree of parallelism in the execution of instructions.
Memory is provided by random access memory (RAM) 402, which in this example is double data rate (DDR) SDRAM totalling 8 gigabytes in capacity. RAM 402 allows storage of frequently-used instructions and data structures by the personal computer 305. A portion of RAM 402 is reserved as shared memory, which allows high speed inter-process communication between applications running on personal computer 305.
Permanent storage is provided by a storage device such a solid-state disk (SSD) 403, which in this instance has a capacity of 512 gigabytes. SSD 403 stores operating system and application data. In alternative embodiments, a hard disk drive could be provided, or several storage devices provided and configured as a RAID array to improve data access times and/or redundancy.
A network interface 404 allows personal computer 305 to connect to a packet-based network such as the Internet. Additionally, personal computer 305 comprises a peripheral interface 405 such as a universal serial bus (USB®). In this embodiment, the peripheral interface 405 connects the personal computer 305 to the camera 304 to allow transfer of the images obtained in use. Whilst the peripheral interface 405 in the present embodiment is a wired USB® connection, it is contemplated that other connection types may be used such as wireless connections using, for example, an 802.11x standard.
Personal computer 305 also comprises an optical drive, such as a CD-ROM drive 406, into which a non-transitory computer readable medium such as an optical disk, e.g. CD-ROM 407 can be inserted. CD-ROM 407 comprises computer-readable instructions to enable the processing method of the present invention. These are, in use, installed on solid-state disk 403, loaded into RAM 402 and then executed by CPU 401. Alternatively, the instructions may be downloaded from a network via the network interface 404 as packet data 408.
It is to be appreciated that the above system is merely an example of a configuration of system that can fulfil the role of personal computer 305. Any other system having a processing device, memory, a storage device and a network interface could equally be used. Thus in an alternative embodiment it is envisaged that an application-specific integrated circuit (ASIC) or field-programmable gate array (FPGA) could be configured with the same instructions so as to perform substantially the same operations as the personal computer 305.
Operations carried out with the measurement system 301 in practice are detailed in
At step 501, the measurement system 301, comprising the laser 302, the camera 304 and personal computer 305, is powered on. At step 502 a question is asked as to whether image processing instructions have been installed on the personal computer 305. If not, then at step 503 the instructions are obtained either from CD-ROM 407 or via the network connection 404 as previously described. Following install, or if the instructions have previously been installed, then at step 504 a manual procedure is carried out consisting of locating a component, such as fan blade 201, in the fixture 302.
The measurement system 301 then captures images at step 505, and processes them at step 506. Procedures carried out to capture images during step 505 will be described with reference to
Following processing of the images, a question is asked at step 507 as to whether another component is to be analysed. If this question is answered in the affirmative, then control returns to step 504 where the new component is located in the fixture 302. If the question asked at step 507 is answered in the negative, then at step 508 the measurement system is shut down.
A procedure carried out during step 505 to obtain the images of the speckle caused by the scattering of the laser light from the surface of the component are set out in
At step 601, a sample location on the surface of the component being imaged is selected. Depending on the size of the component, the width of the beam produced by laser 303. and the requirements of the particular test regime being carried out, it will be appreciated that the sample location chosen may be one of few (perhaps even one), or one of a large number. In this step, the laser 303 and camera 304 may be positioned and angled to illuminate and collect light from the sample location.
At step 602, the camera parameters are selected to produce a correct exposure. In the present example, the camera 304 has a fixed aperture lens and so in this step a metering operation takes place in which exposure length and sensor sensitivity are adjusted to give a correct exposure. Such procedures are known.
At step 603, an exposure is taken of the speckle pattern caused by the scattering of the laser light by the surface at the sample location. The image produced by the camera 304 is then transmitted at step 604 to the personal computer 305, whereupon it may be committed to storage on the solid-state disk 403.
At step 605, a question is asked as to whether any further surface locations are to be sampled. If so, then control returns to step 601 and the above procedures are repeated. If the question asked at step 605 is answered in the negative, then all surface locations have been imaged and step 505 is complete.
An image 701 is shown in
An image 702 is shown in
An image 703 is shown in
As may be discerned from the images 701, 702 and 703, the coarser the surface (the lower grit value in the present example), the lower the contrast in the image. The present invention therefore utilises an image processing scheme to quantify the differences between the speckle patterns caused by surfaces of different roughness, thereby allowing a determination to be made as to the roughness of the surface.
Steps carried out in the present embodiment by the personal computer 305 to process the images in step 506 are set out in
At step 801, an image stored on the solid-state disk 403 at step 604 is loaded into memory for processing. At step 802, the image is converted into a binary image, i.e. pixels are classified as either foreground pixels or background pixels depending upon their level in relation to a threshold. In the present embodiment, the threshold is one of a number of possible candidate thresholds. Selection of the candidate threshold to be used for foreground/background classification may performed unsupervised and automatically. This process will be described further with reference to
Regions of connected components in the binary image are then identified in step 803, which process will be described with reference to
In the present embodiment, a flood-fill approach is taken to identify the regions, but it will be appreciated that any suitable algorithm for identifying connected components may be used. As will be described further with reference to
In this way, an indication of a measure of the surface roughness may be output at step 804 by using a lookup table stored in memory to convert the statistics of the output of step 803 into a measure of roughness, for example. In the present embodiment, the statistics are the total number of regions of connected components and the number of pixels in the largest region of connected components.
Alternatively, a relation may be derived from empirical testing which may be stored as part of the program instructions of the present invention, which is then used during step 804 to convert the statistics generated from the output of step 803 into a measure of surface roughness.
Following production of a measure of the roughness for the surface that caused the speckle pattern which was the subject of the image loaded at step 801, a question is asked at step 805 as to whether any other images are to be processed. If so, control returns to step 801 where the next image is loaded for processing. If not, then the question is answered in the negative and step 506 is complete.
As described previously, in the present embodiment the procedure for conversion into a binary image may be carried out unsupervised and automatically. The processes carried out during step 802 to achieve this are set out in
At step 901, the histogram of the image loaded into memory at step 801 is computed. As will be familiar to those skilled in the art, the histogram of an image comprises bins, each one of which represents a certain level. Each bin records the number of pixels in the image having that particular level.
At step 902, a candidate threshold is selected. For the present case where the images produced by the camera 304 are 8 bit greyscale images, there may be 254 thresholds, corresponding to levels 1 to 254, which would respectively split the image into background pixels of level 0 and foreground pixels of level 1 and over, and background pixels of level 254 and under and foreground pixels of level 255. However, it will be appreciated that any number of thresholds could be provided, in any distribution. Thus for example, there may be 32 candidate thresholds, each separated by 8 levels. Alternatively, the distribution of thresholds may resemble a gamma curve for example, with lower level thresholds being more spaced apart than higher ones, for example.
Thus at step 903 the pixels in the image are classified as either foreground pixels if they are of greater than or equal level to the threshold. or background pixels if they are lower in level than the threshold.
The histogram produced at step 901 is then utilised to perform step 904. In this step, the background pixels are considered a class C0 and the background pixels are another class C1. The probabilities of class occurrence or weights are ω0 and ω1, which are the zeroth order cumulative moments of the histogram either side of the threshold. The class mean levels are μ0 and μ1, which are the first order cumulative moments of the histogram either side of the threshold.
It may be shown that the optimum threshold is located at a point where there is both the lowest variance in the levels of the background pixels and the lowest variance in the levels of the background pixels, i.e. a minimised within-class variance. It may be shown that maximising the between-class variance is equivalent to minimising the within-class variance, and is more computationally efficient.
The between-class variance σ is evaluated by calculating the product of the weights, multiplied by the square of the difference between the background mean and the foreground mean, i.e. ω0ω1(μ0−μ1)2.
At step 905, a question is asked as to whether the value of σ evaluated at step 904 is the maximum thus far. If so, then the candidate threshold selected at the current iteration of step 902 is set as the new optimum threshold. Then, or if the question asked at step 905 was answered in the negative, then a further question is asked as to whether another candidate threshold needs to be considered. If so, control returns to step 902 where the next candidate threshold is selected. If not, then control proceeds to step 908 where a binary image is outputted utilising the optimum threshold.
Binary images derived from the images of the speckle patterns 701, 702 and 703 are shown in
Procedures carried out to identify regions of connected components in the binary image during step 803, are set out in
At step 1101, an iterative variable L is set to an initial value, which in this example is 1. Variable L is used to label pixels as forming part of a particular region of connected components in the image. At step 1102, the first unlabelled foreground pixel (P) in the image is selected, and is labelled with L at step 1103.
A question is then asked at step 1104 as to whether there are any pixels neighbouring the pixel P which are also unlabelled foreground pixels (N).
In the present embodiment, a pixel is a neighbouring pixel N if it is 8-connected to the pixel P. In an alternative implementation, a pixel may be considered a neighbouring pixel if it is only 4-connected to the pixel P. If the question asked at step 1104 is answered in the affirmative, then the labels of the neighbouring pixels N are set to L at step 1105. Control returns to step 1104, where the question asked results in unlabelled foreground pixels which neighbour the original neighbouring pixels (if they exist) also being labelled with L at step 1105. This process iterates until the question asked at step 1104 is answered in the negative. At this point, a region of connected components has been identified, and the pixels in the region have all been labelled with L.
Control therefore proceeds to step 1106, where a question is asked as to whether there is a further unlabelled foreground pixel in the image. If so, then L is incremented at step 1107 and control returns to step 1102 where the next unlabelled foreground pixel is selected, and it, and as a result of the iterative steps 1104 and 1105, the region it forms part of, are labelled with the new value of L.
Eventually, all foreground pixels will be labelled, and step 803 will be complete. Thus, the total number of regions of connected components and the number of pixels in the largest region of connected components may be counted.
A plot of surface roughness—in this case grit number—against the total number of regions of connected components is shown in
As described previously, and as may be seen from the plot, it has been found by the present inventors that there is a relationship between the roughness of a surface, and the number of regions of connected components in the binary images derived from images of the speckle pattern caused by scattering of coherent light from the surface. In this case, there is a positive correlation: the higher the roughness (lower grit number), the higher the number of regions of connected components.
Further, referring now to
As described previously, by utilising the two measures illustrated in
It will be appreciated by those skilled in the art that the conversion to a binary image does impose latency in terms of the processing of the image loaded at step 801. However, the use of a binary image as the basis for identifying regions of connected components means that a significant reduction in processing time at step 803 is achieved.
This is because the binary image simply classifies each pixel as being in one of two states, which greatly reduces the complexity of the algorithm required for identification of connected components. By contrast, greyscale images, where each pixel may occupy one of a plurality of levels (e.g. 256 for an 8 bit image), require for example parallel processing techniques to be employed which inherently are greater in terms of complexity.
Due to the reduction in processing time required by steps 802 and 803 it is envisaged that the present invention may be applied to production-line measurement processes.
It will be understood that the invention is not limited to the embodiments described herein, and various modifications and improvements can be made without departing from the concepts described. Except where mutually exclusive, any of the features may be employed separately or in combination with any other features and the invention extends to and includes all combinations and sub-combinations of one or more features described herein.
| Number | Date | Country | Kind |
|---|---|---|---|
| 1705406.5 | Apr 2017 | GB | national |
