JP5684505B2 - Image processing apparatus, image processing method, and computer program - Google Patents

Description

  The present invention relates to an image processing apparatus and image processing for comparing a multi-value image obtained by imaging a determination target and a distance image obtained by converting a distance from an imaging unit into a gray value with a reference multi-value image and a distance image. The present invention relates to a method and a computer program.

  Conventionally, a method has been developed for determining whether or not a determination target object is a non-defective product by simply comparing a multi-value image acquired by imaging the determination target object with a reference multi-value image. However, by simply comparing the two images, it is not possible to take into account pixel value fluctuations due to shape variations, calculation errors, noise, etc. of the non-defective product itself, and erroneously determine that it is non-defective even though it is non-defective For example, there is a possibility that correct product determination may not be performed correctly.

  In order to solve such a problem, for example, in Patent Document 1, a multi-value image relating to a plurality of non-defective products is prepared, and an average image and a standard deviation image thereof are obtained to enable stable non-defective product determination. Specifically, after aligning multi-valued images related to a plurality of non-defective products, an average value and a standard deviation of pixel values are calculated for each pixel having the same coordinates. After aligning the multi-value image of the judgment object with the multi-value (average) image related to the non-defective product, the difference value from the average value is calculated for each pixel at the same coordinate and defined based on the standard deviation The non-defective product is determined by comparing the threshold image for each pixel.

  In this way, pixel value fluctuations due to non-defective product shape variations, calculation errors, noise, etc. can be effectively eliminated according to the degree of pixel value variation for each pixel, and non-defective products can be obtained for each pixel. Since the range to be determined is different, an appropriate non-defective threshold value can be set for each non-defective part. Therefore, the non-defective product can be determined with higher accuracy.

JP 2005-265661 A

  However, the image processing method disclosed in Patent Document 1 cannot detect unevenness on the surface of the determination object that is difficult to appear as a difference in light and shade values. Therefore, even though the difference in color and unevenness in the surface are included in the non-defective product distribution range, the way the light hits changes due to the variation in the position where the judgment object is placed, etc. May be erroneously determined to be non-defective due to being detected larger than actual.

  It is also conceivable to compare the three-dimensional data obtained by measuring the three-dimensional shape of the determination target with the three-dimensional CAD data of the determination target in order to determine the difference in the three-dimensional shape. However, it is necessary to acquire accurate three-dimensional data as measured values, and it takes a considerable amount of time to determine whether or not the product is non-defective, such as configuring the three-dimensional data with a plurality of measurement data measured from many directions. And the device itself is expensive.

  Furthermore, when a defect occurs in the vicinity of the contour line, it is difficult to recognize the defect as the standard deviation of the pixel value is large in the vicinity of the contour line, and the sensitivity to detect the presence of the defect is low. There was a problem. The increase in the standard deviation of the pixel value in the vicinity of the contour line is caused by the variation in the position where the determination target is placed, the difference in the imaging position, the shape variation of the non-defective product itself, and the like.

The present invention has been made in view of the above circumstances, and an appropriate feature amount can be selected according to the properties of the second multi-valued image or the second distance image. An object of the present invention is to provide an image processing apparatus, an image processing method, and a computer program that can detect the difference between the two with high sensitivity .

In order to achieve the above object, an image processing apparatus according to a first aspect of the present invention is an image processing apparatus for performing non-defective product determination by comparing a multi-value image obtained by capturing an object to be determined with a multi-value image group related to non-defective products. A first non-defective image acquisition unit that acquires a plurality of first multi-valued images related to the acquired non-defective product, and a first value obtained by converting the distance from the imaging unit into a gray value for each pixel of the acquired first multi-level image. A non-defective distance image generating unit that generates a distance image of the first multi-value image, a first feature amount extracting unit that extracts a pixel value related to a color component as a feature amount for each pixel of the first multi-value image, and the first multi-value image At least one feature among a plurality of feature quantity extraction means including at least a second feature quantity extraction means for extracting edge strengths in two different directions as feature quantities for each pixel of the value image or the first distance image Selection of quantity extraction means A selection accepting means for accepting the acquired first gray value of at least every one of the pixels of the multivalued image and the generated the first distance image, the distribution range calculation means for calculating a distribution range for performing good decision A second multi-valued image obtaining unit for obtaining a second multi-valued image relating to the determination target, and a second value obtained by converting the distance from the imaging unit into a gray value for each pixel of the obtained second multi-valued image. Whether or not the target distance image generating means for generating the distance image, the acquired second multi-value image, and the gray value for each pixel of the generated second distance image are included in the corresponding distribution range. Determination means for determining whether or not the distribution range calculation means selects the first distribution range when the selection of the first feature quantity extraction means is accepted, and selects the second feature quantity extraction means. If you accept the second The distribution range is calculated based on the feature amount extracted by the feature amount extraction unit that has received the selection, and the determination unit is configured for each pixel of the second multi-value image or the second distance image. In addition, a feature amount is extracted by the feature amount extraction unit that has received the selection, and it is determined whether the feature amount is included in the first distribution range or the second distribution range corresponding to the extracted feature amount. It is characterized by being.

  An image processing apparatus according to a second aspect of the present invention includes the plurality of imaging units in the first invention, and the non-defective distance image generation unit or the target distance image generation unit includes a plurality of images captured by the plurality of imaging units. The first distance image or the second distance image is generated based on the first multi-value image or the second multi-value image.

  An image processing apparatus according to a third aspect of the present invention is the image processing apparatus according to the first aspect, further comprising the imaging unit and a light irradiation unit that irradiates a determination target with light, and the good product distance image generation unit or the target distance image. The generation unit generates the first distance image or the second distance image based on the first multi-value image or the second multi-value image captured by the imaging unit and the irradiated light. It is characterized by the above.

The image processing apparatus according to a fourth aspect of the present invention is the image processing apparatus according to any one of the first to third aspects, wherein when the selection of the first feature amount extraction unit is received, the distribution range calculation unit Pixel average value calculating means for calculating an average value of pixel values relating to color components for each pixel of the multi-value image, and pixels of the first multi-value image based on the calculated average values of pixel values relating to color components Each having a pixel value distribution range calculation means for calculating the first distribution range of pixel values in a multi-dimensional space having each color component as a coordinate axis, and the determination means includes the second multi-value image. For each pixel, a pixel value related to the color component is calculated, and it is determined whether or not the pixel value is included in the first distribution range of the pixel value in the multidimensional space with each color component as a coordinate axis. Receives selection of second feature extraction means When attached, the distribution range calculation means includes edge strength calculation means for each direction for calculating edge strength in two different directions for each pixel of the first multi-value image or the first distance image, and the calculated edge Based on the intensity, edge intensity average value calculating means for calculating an average value of edge intensity in two different directions for each pixel of the first multi-value image or the first distance image, and centering the calculated average value An edge intensity distribution range calculation means for calculating a cross-correlation distribution range of edge strengths in two different directions as the second distribution range for each pixel of the first multi-valued image or the first distance image. And the determination means calculates edge strengths in two different directions for each pixel of the second multi-valued image or the second distance image, and the calculated edge strengths correspond to the corresponding phases. Characterized in that you have to determine whether or not included in the correlation distribution range.

An image processing apparatus according to a fifth aspect of the present invention is the image processing apparatus according to the fourth aspect , wherein the direction-specific edge strength calculating means calculates the edge strength in two directions orthogonal to each other.

The image processing apparatus according to a sixth aspect of the present invention is the image processing apparatus according to the fifth aspect , wherein the direction-specific edge strength calculating means calculates the edge strength in two directions of a row direction and a column direction in a two-dimensional image. It is characterized by being.

An image processing apparatus according to a seventh invention is the image processing apparatus according to any one of the fourth to sixth inventions, wherein the edge intensity distribution range calculation means sets the cross-correlation distribution range as the virtual ellipse region to the second It is characterized in that it is calculated as a distribution range .

According to an eighth aspect of the present invention, in any one of the first to seventh aspects, the first multi-value image and the second multi-value image are aligned or the first distance is set. Position adjustment means for aligning the image with the second distance image is provided.

An image processing apparatus according to a ninth invention is the image processing apparatus according to any one of the first to eighth inventions, wherein the selection accepting means accepts selection of a plurality of feature quantity extracting means. However, any one of the feature quantities extracted by the feature quantity extraction unit that has received the selection is not included in the first distribution range or the second distribution range corresponding to each feature quantity. When the determination is made, it is determined that the determination object is not a non-defective product.

Next, in order to achieve the above object, the image processing method according to the tenth aspect of the present invention is executed by an image processing apparatus that performs a non-defective product determination by comparing a multi-value image obtained by imaging a determination target with a multi-value image group related to a good product In the image processing method that can be performed, a step of acquiring a plurality of first multi-value images related to non-defective products imaged by the imaging unit, and for each pixel of the acquired first multi-value image, from the imaging unit Generating a first distance image obtained by converting the distance into a gray value, and a first feature amount extraction means for extracting a pixel value related to a color component as a feature amount for each pixel of the first multi-value image; A plurality of feature quantity extraction means including at least a second feature quantity extraction means for extracting edge strengths in two different directions as feature quantities for each pixel of the first multi-value image or the first distance image. At least 1 Calculating a step of accepting a selection of a feature extraction means, the gray value of each at least one of the pixels of the first multivalued image, and generated the first distance image acquired, the distribution range for performing good determination of A second distance image obtained by converting a distance from the imaging unit into a gray value for each pixel of the acquired second multi-value image And determining whether or not a gray value for each pixel of at least one pixel of the acquired second multi-valued image and the generated second distance image is included in the corresponding distribution range; Including a first distribution range when the selection of the first feature amount extraction unit is received, and a second distribution range when the selection of the second feature amount extraction unit is received, respectively. Accept selection Calculated based on the feature amount extracted by the feature amount extraction means, and for each pixel of the second multi-valued image or the second distance image, the feature amount extraction means that has received a selection extracts the feature amount, It is characterized by determining whether it is included in said 1st distribution range or said 2nd distribution range corresponding to the extracted feature-value.

Next, in order to achieve the above object, a computer program according to an eleventh aspect of the invention executes an image processing apparatus that performs a non-defective product determination by comparing a multi-value image obtained by imaging a determination target with a multi-value image group related to a non-defective product. A non-defective image acquiring unit that acquires a plurality of first multi-valued images related to a non-defective product that has been imaged by the imaging unit in the computer program capable of being acquired, for each pixel of the acquired first multi-valued image In addition, a non-defective distance image generation unit that generates a first distance image obtained by converting the distance from the imaging unit into a gray value, and extracts a pixel value related to a color component as a feature amount for each pixel of the first multi-value image First feature quantity extracting means, and second feature quantity extracting means for extracting edge strengths in two different directions as feature quantities for each pixel of the first multi-value image or the first distance image. Among the plurality of feature extraction means including at least a selection accepting means for accepting a selection of at least one feature extraction means, for each of at least one of the pixels of the first multivalued image, and generated the first distance image acquired Distribution range calculation means for calculating a distribution range for performing non-defective product determination of gray value, multi-value image acquisition means for acquiring a second multi-value image relating to the determination target, and for each pixel of the acquired second multi-value image In addition, target distance image generation means for generating a second distance image obtained by converting the distance from the imaging means into a gray value, and at least one of the acquired second multi-value image and the generated second distance image It functions as a determination unit that determines whether or not the gray value for each pixel is included in the corresponding distribution range, and the distribution range calculation unit accepts the selection of the first feature amount extraction unit If the selection of the second feature amount extraction unit is accepted, the second distribution range is selected based on the feature amount extracted by the feature amount extraction unit that has accepted the selection. It is made to function as a means for calculating, and the determination means extracts the feature amount by the feature amount extraction means that accepts selection for each pixel of the second multi-value image or the second distance image, and the extracted feature It is made to function as a means to judge whether it is contained in said 1st distribution range or said 2nd distribution range corresponding to quantity.

In the first invention, the tenth invention, and the eleventh invention, a plurality of first multi-value images related to non-defective products, which are imaged by the imaging means, are acquired, and the pixels of the acquired first multi-value image are obtained from the imaging means. A first distance image is generated by converting the distance to a gray value. The first feature value extraction means for extracting the pixel value related to the color component as a feature value for each pixel of the first multivalued image and two different pixels for each pixel of the first multivalued image or the first distance image. The selection of at least one feature quantity extraction means is accepted from among a plurality of feature quantity extraction means including at least a second feature quantity extraction means for extracting edge strength in the direction as a feature quantity. A distribution range for performing non-defective product determination of the gray value for each pixel of at least one of the acquired first multi-value image and the generated first distance image is calculated. A second multi-value image related to the determination target is acquired, and for each pixel of the acquired second multi-value image, a second distance image in which the distance from the imaging unit is converted into a gray value is generated and acquired. It is determined whether the gray value for each pixel of at least one of the second multi-valued image and the generated second distance image is included in the corresponding distribution range. Since it is determined based on the second distance image whether or not it is a non-defective product, it is possible to detect unevenness on the surface of the determination object that is difficult to detect by the conventional method, and to place the determination object Even if the way the light strikes changes due to variations in the position, etc., it is possible to reduce the possibility of erroneously determining that the light is not a good product. When the selection of the first feature quantity extraction unit is accepted, the first distribution range is accepted, and when the selection of the second feature quantity extraction unit is accepted, the second distribution range is accepted. Calculated based on the feature quantity extracted by the feature quantity extraction means, and extracted for each pixel of the second multi-valued image or the second distance image by extracting the feature quantity by the feature quantity extraction means that accepted the selection. Since it is determined whether or not it is included in the first distribution range or the second distribution range corresponding to the feature amount, the feature amount extraction is performed according to the property of the second multi-value image or the second distance image. By selecting an appropriate feature amount, it is possible to detect the presence of a defect even when the defect occurs in the vicinity of the contour line, and with high sensitivity for color differences. It becomes possible to detect. For example, when the feature amount that has received the selection is a pixel value related to a color component, it is possible to detect the presence or absence of shading variation due to a difference in color, and when the feature amount that has received the selection is an edge strength in two different directions, The presence of a defect near the contour line can be detected.

  In the second invention, the first distance image or the second distance image is provided on the basis of the plurality of first multi-value images or second multi-value images that are provided with a plurality of image pickup means and are imaged by the plurality of image pickup means. Generate. By providing a plurality of imaging means, a plurality of multi-value images can be acquired, and a distance image can be generated by a so-called passive method such as a stereo measurement method.

  In the third invention, the first multi-valued image or the second multi-valued image picked up by the image pickup means and the light irradiation means for irradiating the determination target with light, and the irradiated light The first distance image or the second distance image is generated based on the above. Accordingly, light can be actively irradiated to measure the distance in the height direction, and a distance image can be generated by a so-called active method such as a light cutting method or a pattern projection method.

In the fourth aspect of the invention, when the selection of the first feature amount extraction unit is accepted, the average value of the pixel values related to the color component is calculated for each pixel of the first multi-value image, and the pixel value related to the calculated color component is calculated. Based on the average value, a first distribution range of pixel values in a multidimensional space with each color component as a coordinate axis is calculated for each pixel of the first multivalued image. For each pixel of the second multi-valued image, a pixel value related to the color component is calculated, and it is determined whether or not the pixel value is included in the first distribution range of the pixel value in the multi-dimensional space with each color component as a coordinate axis. . When the selection of the second feature amount extraction means is accepted, the edge strength in two different directions is calculated for each pixel of the first multi-value image or the first distance image, and the first edge value is calculated based on the calculated edge strength. For each pixel of one multi-value image or first distance image, an average value of edge strengths in two different directions is calculated, and the first multi-value image or first distance image is calculated around the calculated average value. For each pixel, a cross-correlation distribution range of edge strengths in two different directions is calculated as a second distribution range . For each pixel of the second multi-value image or the second distance image, the edge strength in two different directions is calculated, and it is determined whether or not the calculated edge strength is included in the corresponding cross-correlation distribution range. As a result, when the feature quantity accepted for selection is a pixel value related to a color component, it is possible to detect a color difference depending on whether the calculated pixel value is included in the first distribution range, and accept the selection. If the detected feature amounts are edge strengths in two different directions, it is possible to detect the presence of a defect near the contour line depending on whether or not it is included in the cross-correlation distribution range of the calculated edge strengths.

In the fifth invention, by calculating the edge strength in two directions orthogonal to each other, the calculation of the cross-correlation distribution range is facilitated, and the processing load can be reduced.

In the sixth invention, by calculating the edge strength in two directions of the row direction and the column direction in the two-dimensional image, the calculation of the cross-correlation distribution range is further facilitated, and the processing load can be further reduced. Become.

In the seventh invention, by calculating the cross-correlation distribution range as the second distribution range as a virtual elliptical region, edges in different two directions calculated for each pixel of the second multi-valued image or the second distance image It is possible to more easily determine whether or not the intensity is included in the cross-correlation distribution range.

In the eighth invention, the second multi-value image or the second multi-value image is aligned by aligning the first multi-value image and the second multi-value image, or the first distance image and the second distance image. It is possible to more accurately determine whether or not the edge intensity in two different directions calculated for each pixel of the second distance image is included in the cross-correlation distribution range.

In the ninth invention, when selection of a plurality of feature quantity extraction means is accepted, one of the feature quantities extracted by the feature quantity extraction means that has accepted the selection corresponds to each feature quantity. When it is determined that it is not included in the first distribution range or the second distribution range , it is determined that the determination target object is not a non-defective product. As a result, even if it was determined that a non-defective product was determined to be a non-defective product based on an inappropriate feature value, it was correctly determined that it was not a non-defective product based on an appropriate feature value. It becomes possible to judge.

According to the present invention, since it is determined whether or not it is a non-defective product based on the second distance image, it is possible to detect unevenness on the surface of the determination target that is difficult to detect by the conventional method, Even if the way the light hits changes due to variations in the position where the determination target is placed, it is possible to reduce the possibility that it is erroneously determined that it is not a non-defective product. In addition, the feature quantity extraction unit can be selected according to the properties of the second multi-value image or the second distance image, and defects are generated in the vicinity of the contour line by selecting an appropriate feature quantity. Even in this case, the presence of a defect can be detected, and a difference in color can be detected with high sensitivity. For example, when the feature amount that has received the selection is a pixel value related to a color component, it is possible to detect the presence or absence of shading variation due to a difference in color, and when the feature amount that has received the selection is an edge strength in two different directions, The presence of a defect near the contour line can be detected.

1 is a block diagram schematically showing a configuration of an image processing apparatus according to Embodiment 1 of the present invention. It is a functional block diagram which shows the example of 1 structure of the image processing apparatus which concerns on Embodiment 1 of this invention. It is the perspective view for demonstrating the distance image which concerns on Embodiment 1 of this invention, and the illustration figure of a distance image. It is a flowchart which shows the procedure of the quality information collection process of the main-control part of the image processing apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the procedure of the quality determination process of the main-control part of the image processing apparatus which concerns on Embodiment 1 of this invention. It is a functional block diagram which shows the example of 1 structure of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows the procedure of the quality information collection process of the main control part of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows the procedure of the quality determination process of the main-control part of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a functional block diagram of the distribution range calculation means when the selection of the first feature quantity extraction means of the image processing apparatus according to Embodiment 2 of the present invention is accepted. It is an illustration figure which shows the distribution range of a pixel value when selection of the 1st feature-value extraction means of the image processing apparatus which concerns on Embodiment 2 of this invention is received. It is a flowchart which shows the procedure of the distribution range calculation process at the time of receiving selection of the 1st feature-value extraction means of the main control part of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a functional block diagram of the distribution range calculation means when the selection of the second feature quantity extraction means of the image processing apparatus according to Embodiment 2 of the present invention is accepted. It is an illustration figure of the cross correlation distribution range of the edge intensity | strength in two different directions for every predetermined | prescribed pixel of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows the procedure of the distribution range calculation process at the time of receiving selection of the 2nd feature-value extraction means of the main control part of the image processing apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows the procedure of the quality determination process in case the selection of the 2nd feature-value extraction means of the main control part of the image processing apparatus which concerns on Embodiment 2 of this invention is received.

  Hereinafter, an image processing apparatus according to an embodiment of the present invention will be described with reference to the drawings. Throughout the drawings to be referred, elements having the same or similar configuration or function are denoted by the same or similar reference numerals, and detailed description thereof is omitted.

(Embodiment 1)
FIG. 1 is a block diagram schematically showing a configuration of an image processing apparatus according to Embodiment 1 of the present invention. As shown in FIG. 1, the image processing apparatus 2 according to the first embodiment includes a camera 1 that is an imaging unit that captures a multi-valued image of a determination target object or a non-defective product, and a light that irradiates the determination target object or the non-defective product with laser light. 9 and a captured multi-value image, a generated distance image, or a display device 3 which is an image display means for displaying an image generated in the course of calculation processing.

  The image processing apparatus 2 includes at least a CPU (central processing unit), a main control unit 21 configured by an LSI, a memory 22, a storage unit 23, an input unit 24, an output unit 25, a communication unit 26, an auxiliary storage unit 27, and the above-described unit. The internal bus 28 is connected to the hardware. The main control unit 21 is connected to each hardware unit as described above of the image processing apparatus 2 via the internal bus 28, and controls the operation of each hardware unit described above and is stored in the storage unit 23. Various software functions are executed in accordance with the computer program 5. The memory 22 is composed of a volatile memory such as SRAM or SDRAM, and a load module is expanded when the computer program 5 is executed, and stores temporary data generated when the computer program 5 is executed.

  The storage means 23 is composed of a built-in fixed storage device (hard disk, flash memory), ROM or the like. The computer program 5 stored in the storage means 23 is downloaded by the auxiliary storage means 27 from the portable recording medium 4 such as a DVD, CD-ROM, or flash memory in which information such as programs and data is recorded, and stored at the time of execution. It is expanded from the means 23 to the memory 22 and executed. Of course, a computer program downloaded from an external computer via the communication means 26 may be used.

  The storage unit 23 stores the multi-value image data related to the acquired multi-value images related to the multi-value images, and the multi-value images related to the non-defective products stored in the non-defective product image data storage unit 231. A distribution range information storage unit 232 that stores information (distribution range information) regarding the distribution range of the gray value for each pixel based on the data. The determination target is a non-defective product according to the determination as to whether or not the feature amount of the determination target is included in the distribution range specified based on the distribution range information stored in the distribution range information storage unit 232. It is determined whether or not.

  The communication means 26 is connected to an internal bus 28, and is connected to an external network such as the Internet, a LAN, or a WAN, so that data can be transmitted / received to / from an external computer or the like. In other words, the storage unit 23 described above is not limited to the configuration built in the image processing apparatus 2, but an external device such as a hard disk installed in an external server computer connected via the communication unit 26. It may be a recording medium.

  The input means 24 is a broad concept including all devices for acquiring input information such as a touch panel integrated with a liquid crystal panel or the like in addition to a data input medium such as a keyboard and a mouse. The output means 25 means a printing device such as a laser printer or a dot printer.

  The camera (imaging means) 1 is a CCD camera or the like provided with a CCD image sensor. The display device 3 is a display device having a CRT, a liquid crystal panel, and the like. The light (light irradiating means) 9 can irradiate the determination target object and the non-defective product with laser light, and can acquire a plurality of multi-value images captured by shifting the phase of the sinusoidal fringe pattern, for example.

  The configuration is not limited to using one camera 1 and one light 9, and a plurality of cameras 1 may be used. By using a stereo measurement method or the like based on a plurality of multi-value images acquired by a plurality of cameras 1, 1,..., The distance from the camera 1 to the determination target can be obtained and a distance image can be generated. Because.

  The camera 1, the display device 3, the light 9, and the like may be integrated with the image processing device 2 or may be separated.

  The external control device 6 is a control device connected via the communication means 26, and corresponds to, for example, a PLC (programmable logic controller). Here, the external control device 6 means all devices that perform post-processing according to the image processing result by the image processing apparatus 2.

  FIG. 2 is a functional block diagram showing a configuration example of the image processing apparatus 2 according to Embodiment 1 of the present invention. In FIG. 2, the image processing apparatus 2 according to the first embodiment includes a camera 1, a light 9, an image processing unit 7 that executes processing of the image processing apparatus 2, a storage unit 23, and an image display unit 8. Consists of

  The camera 1 is, for example, a digital camera, captures, for example, a film surface as a determination target, acquires a multi-valued image, and outputs it to the image processing unit 7. Moreover, by irradiating light from the light 9, a fringe pattern is projected on the surface of the determination object or a non-defective product. The camera 1 images the projected fringe pattern. The image processing device 2 acquires the fringe pattern captured by the camera 1 as a fringe pattern image, generates a distance image by executing phase analysis of the fringe pattern, and stores the distance image in the storage device 23. The light irradiation control of the light 9 can be performed by the main control unit 21, for example.

  The image processing unit 7 includes a non-defective image acquisition unit 71, a non-defective product distance image generation unit 72, a distribution range calculation unit 73, a multi-value image acquisition unit 74, a target distance image generation unit 75, a position adjustment unit 76, Determination means 77. The image processing unit 7 includes a main control unit 21, a memory 22, an external I / F, and the like, and includes a non-defective image acquisition unit 71, a non-defective product distance image generation unit 72, a distribution range calculation unit 73, and a multi-value image acquisition. The processing operations of the means 74, the target distance image generating means 75, the position adjusting means 76, and the judging means 77 are controlled.

  The storage unit 23 functions as an image memory, and is based on multi-value image data related to the multi-value image captured by the camera 1 and a fringe pattern image captured by the camera 1 by projecting a stripe pattern projected by irradiating light from the light 9. The distance image data relating to the distance image generated in this way and the image data after performing various processes such as alignment and average value calculation in the image processing unit 7 are stored as needed. You may memorize | store as pixel value data for every pixel instead of as image data.

  The image display unit 8 includes a display device 3 such as a computer monitor. The image display unit 8 displays on the display screen of the display device 3 the determination result as to whether or not the multi-valued image obtained by imaging the determination target that is the target of the non-defective determination and the non-defective product. That is, the multi-value image display unit 81 displays a multi-value image corresponding to the instruction of the image processing unit 7 on the display screen of the display device 3, and the determination result display unit 82 determines whether or not the determination target object is a non-defective product. Are displayed on the display screen of the display device 3.

  Next, each configuration of the image processing unit 7 will be described.

  The non-defective product image acquisition unit 71 acquires a plurality of multi-value images (first multi-valued images) related to the non-defective product imaged by the camera 1. That is, by acquiring a plurality of multi-value images related to non-defective products, a plurality of feature amounts can be extracted for each pixel of the acquired multi-value image. Multi-value image data relating to the acquired multi-value image is stored in the non-defective image data storage unit 231 of the storage means 23.

  The non-defective product distance image generation means 72 performs the phase analysis of the fringe pattern based on the fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1, thereby A non-defective distance image (first distance image) in which the gray value is specified for each pixel according to the distance to the camera 1 is generated. FIG. 3 is a perspective view for explaining the distance image according to the first embodiment of the present invention and an exemplary view of the distance image.

  FIG. 3A is a perspective view for explaining a distance image according to Embodiment 1 of the present invention. In the example of FIG. 3A, the camera 1 is installed above a workpiece 301 having a shape in which two cylinders having different radii are overlapped. The distance between the top surface S1 of the work 301 and the camera 1 is L1, the distance between the middle surface S2 of the work 301 and the camera 1 is L2, and the distance between the mounting surface S3 of the work 301 and the camera 1 is L3. Let H be (L3-L1).

  FIG.3 (b) is an illustration figure of the distance image which concerns on Embodiment 1 of this invention. As shown in FIG. 3B, the uppermost surface S1 of the work 301 is the darkest (for example, black), the mounting surface S3 of the work 301 is the thinnest (for example, white), and the middle surface S2 of the work 301 is an intermediate color (for example, gray) ). That is, according to the distances L1, L2, and L3 between the workpiece 301 and the camera 1, a distance image in which the gray value of each pixel decreases in the order of the top surface S1, the middle surface S2, and the placement surface S3 may be generated. it can.

  The generation method of the distance image is roughly classified into two methods. One is a passive method that uses images taken without using special light irradiation means to acquire images, and the other is active irradiation with light to measure the distance in the height direction. It is an active method to generate. In the first embodiment, an active method is used in which light is emitted from the light 9 and generated.

  The stereo measurement method using two cameras is a representative example of the passive method. In the stereo measurement method, a distance image can be generated simply by placing two cameras at a predetermined position.

  On the other hand, in the active method, a light cutting method and a pattern projection method are given as typical examples. The light cutting method uses a single camera and a single light (light irradiation means) to irradiate a workpiece with a linear laser beam and distort the light image according to the shape of the workpiece surface. To restore the three-dimensional shape of the workpiece.

  The pattern projection method restores the three-dimensional shape of the workpiece by analyzing a plurality of multi-valued images captured by shifting the shape, phase, and the like of a predetermined pattern (for example, a fringe pattern) projected onto the workpiece. In the first embodiment, the phase of the sine wave fringe pattern is shifted to capture at least three multivalued images, the phase of the sine wave is obtained for each pixel, and the phase shift method for obtaining the three-dimensional shape of the workpiece is used. Yes.

  Of course, the present invention is not limited to this, and a moire-potography method is used that restores the three-dimensional shape of the workpiece by using a kind of spatial frequency beat phenomenon that occurs when two regular patterns are synthesized. Alternatively, an optical radar method, a focusing method, a confocal method, a white light interference method, or the like may be used.

  Returning to FIG. 2, the distribution range calculation means 73 calculates a distribution range for performing non-defective product determination of the gray value for each pixel of the generated distance image. Information on the calculated distribution range is stored in the distribution range information storage unit 232 of the storage unit 23 as a function expression indicating the boundary of the distribution range, coordinate values, threshold values, and the like.

  The multi-value image acquisition unit 74 acquires a multi-value image (second multi-value image) related to the determination target imaged by the camera 1. The target distance image generation means 75 performs the phase analysis of the fringe pattern based on the fringe pattern image obtained by capturing the fringe pattern projected by irradiating light from the light 9 with the camera 1, thereby A target distance image (second distance image) in which the gray value for each pixel is specified according to the distance from the surface to the camera 1 is generated. The method for generating a non-defective distance image is the same as that of the non-defective distance image generating means 72.

  The position adjusting unit 76 performs alignment between the target distance image related to the generated determination target object and the non-defective distance image related to the non-defective product. Specifically, an average image of a plurality of non-defective product distance images relating to non-defective products is calculated, and alignment with the average image is performed. The means for aligning the distance images is not particularly limited as long as it is a known technique. For example, the positions of feature points of both distance images may be detected by pattern matching, and alignment may be performed using an upper limit value and a lower limit value that can be determined to match. Further, the degree of coincidence between both distance images may be calculated by calculating a normalized correlation or the like, and alignment may be performed so that the degree of coincidence is greater than a predetermined value. Of course, the alignment may be performed so that the contour line, area, center of gravity, etc. of both distance images coincide.

  In the present embodiment, the position adjustment unit 76 performs alignment between the target distance image related to the generated determination target object and the non-defective product distance image related to the non-defective product, but is not particularly limited thereto. For example, when calculating an average image of a plurality of non-defective distance images regarding non-defective products, the alignment may be performed between the non-defective item distance images. When performing an alignment between a plurality of non-defective distance images when calculating an average image of a plurality of non-defective distance images regarding the non-defective products, the alignment may be performed by the position adjusting means 76, or the position adjustment between the non-defective distance images may be performed separately. Means and the like may be provided.

  The determination unit 77 determines whether the gray value for each pixel of the target distance image related to the generated determination target object is included in the distribution range of the gray value for each pixel of the non-defective distance image calculated for the good product. . If it is determined that the gray value for each pixel of the target distance image related to the determination target is included in the distribution range, the determination target is determined to be non-defective, and if it is determined that it is not included, It is determined that the determination target is not a non-defective product.

  FIG. 4 is a flowchart showing a non-defective information collection process procedure of the main control unit 21 of the image processing apparatus 2 according to the first embodiment of the present invention. The main control unit 21 of the image processing apparatus 2 obtains a plurality of multi-value images (first multi-value images) related to non-defective products captured by the camera 1 (step S401). The main control unit 21 stores the acquired multi-value image data of the multi-value image group in the non-defective product image data storage unit 231 of the storage unit 23 (step S402).

  The main control unit 21 obtains a fringe pattern image obtained by photographing the fringe pattern projected by irradiating light from the light 9 with the camera 1 for a plurality of non-defective products (step S403), and executes the phase analysis of the fringe pattern. Thus, a plurality of non-defective product distance images in which the gray value is specified for each pixel according to the distance from the non-defective product surface to the camera 1 are generated (step S404). The main control unit 21 stores the generated distance image data of the non-defective distance image group in the storage unit 23 (step S405).

  The main control unit 21 calculates a distribution range for performing non-defective product determination based on the generated non-defective product distance image (step S406). The main control unit 21 stores the information on the calculated distribution range in the distribution range information storage unit 232 of the storage unit 23 as a function expression indicating the boundary of the distribution range, coordinate values, threshold values, and the like (step S407). In addition, when calculating the distribution range for performing non-defective product determination, alignment may be performed between a plurality of non-defective product distance images.

  FIG. 5 is a flowchart showing a non-defective product determination process performed by the main control unit 21 of the image processing apparatus 2 according to the first embodiment of the present invention. The main control unit 21 of the image processing apparatus 2 acquires a multi-value image (second multi-value image) related to the determination target imaged by the camera 1 (step S501). The main control unit 21 obtains a fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1 (step S502), and performs a phase analysis of the fringe pattern to thereby determine a determination target. A target distance image specifying a gray value for each pixel according to the distance from the surface of the object to the camera 1 is generated (step S503).

  The main control unit 21 aligns the target distance image related to the generated determination target object and the non-defective product distance image related to the non-defective product (Step S504). Specifically, an average distance image of a plurality of non-defective product distance images related to the non-defective product is calculated, and alignment is performed with the average distance image. The means for aligning the distance images is not particularly limited as long as it is a known technique. For example, the positions of feature points of both distance images may be detected by pattern matching, and alignment may be performed using an upper limit value and a lower limit value that can be determined to match. Further, the degree of coincidence between both distance images may be calculated by calculating a normalized correlation or the like, and alignment may be performed so that the degree of coincidence is greater than a predetermined value. Of course, the alignment may be performed so that the contour line, area, center of gravity, etc. of both distance images coincide.

  The main control unit 21 determines whether or not the gray value for each pixel of the target distance image regarding the generated determination target object is included in the distribution range of the gray value for each pixel of the non-defective distance image regarding the calculated good product. (Step S505). When the main control unit 21 determines that the grayscale value for each pixel of the target distance image related to the determination target object is included in the distribution range (step S505: YES), the main control unit 21 determines that the determination target object is a non-defective product. It is determined that there is, and information indicating that the determination target is a non-defective product is displayed on the display screen of the display device 3 as a determination result (step S506).

  When the main control unit 21 determines that the grayscale value for each pixel of the target distance image related to the determination target is not included in the distribution range (step S505: NO), the main control unit 21 determines that the determination target is a non-defective product. The information indicating that the determination target is not a non-defective product is displayed on the display screen of the display device 3 as a determination result (step S507).

  As described above, according to the first embodiment, since it is determined whether or not it is a non-defective product based on the second distance image, the unevenness on the surface of the determination target that is difficult to detect by the conventional method. It is possible to reduce the possibility of erroneously determining that the product is not a good product even if the way the light hits changes due to variations in the position where the determination target is placed, etc. Become.

(Embodiment 2)
Since the configuration of the image processing apparatus according to the second embodiment of the present invention is the same as that of the first embodiment, detailed description thereof is omitted by attaching the same reference numerals. The second embodiment has a plurality of feature amount extraction means, and it is implemented by determining whether or not it is a non-defective product by extracting the feature amounts of each of the acquired multi-value image or generated distance image. This is different from the first embodiment.

  FIG. 6 is a functional block diagram showing a configuration example of the image processing apparatus 2 according to Embodiment 2 of the present invention. In FIG. 6, the image processing apparatus 2 according to the second embodiment is similar to the first embodiment in that the camera 1, the light 9, the image processing unit 7 that executes the processing of the image processing apparatus 2, and the storage unit 23. And an image display unit 8.

  The camera 1 is, for example, a digital camera, captures, for example, a film surface as a determination target, acquires a multi-valued image, and outputs it to the image processing unit 7. Moreover, by irradiating light from the light 9, a fringe pattern is projected on the surface of the determination object or a non-defective product. The camera 1 images the projected fringe pattern. The image processing device 2 acquires the fringe pattern captured by the camera 1 as a fringe pattern image, generates a distance image by executing phase analysis of the fringe pattern, and stores the distance image in the storage device 23.

  The image processing unit 7 includes a non-defective image acquiring unit 71, a non-defective product distance image generating unit 72, a selection receiving unit 78, a feature amount extracting unit 79 (selected feature amount extracting unit 79c), a distribution range calculating unit 73, A value image acquisition unit 74, a target distance image generation unit 75, a position adjustment unit 76, and a determination unit 77 are included. The image processing unit 7 includes a main control unit 21, a memory 22, an external I / F, and the like, and includes a non-defective image obtaining unit 71, a non-defective product distance image generating unit 72, a selection receiving unit 78, and a feature amount extracting unit 79. (Selected feature amount extraction means 79c), distribution range calculation means 73, multi-value image acquisition means 74, target distance image generation means 75, position adjustment means 76, and determination means 77 are controlled.

  The storage unit 23 functions as an image memory, and is based on multi-value image data related to the multi-value image captured by the camera 1 and a fringe pattern image captured by the camera 1 with a stripe pattern projected by irradiating light from the light 9. The distance image data relating to the distance image generated in this way and the image data after performing various processes such as alignment and average value calculation in the image processing unit 7 are stored as needed. You may memorize | store as pixel value data for every pixel instead of as image data.

  The image display unit 8 includes a display device 3 such as a computer monitor. The image display unit 8 displays on the display screen of the display device 3 the determination result as to whether or not the multi-valued image obtained by imaging the determination target that is the target of the non-defective determination and the non-defective product. That is, the multi-value image display unit 81 displays a multi-value image corresponding to the instruction of the image processing unit 7 on the display screen of the display device 3, and the determination result display unit 82 determines whether or not the determination target object is a non-defective product. Are displayed on the display screen of the display device 3.

  Next, each configuration of the image processing unit 7 will be described.

  The non-defective product image acquisition unit 71 acquires a plurality of multi-value images (first multi-valued images) related to the non-defective product imaged by the camera 1. That is, by acquiring a plurality of multi-value images related to non-defective products, a plurality of feature amounts can be extracted for each pixel of the acquired multi-value image. Multi-value image data relating to the acquired multi-value image is stored in the non-defective image data storage unit 231 of the storage means 23.

  The non-defective product distance image generation means 72 performs the phase analysis of the fringe pattern based on the fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1, thereby A non-defective distance image (first distance image) in which the gray value is specified for each pixel according to the distance to the camera 1 is generated.

  The selection receiving unit 78 receives selection of at least one feature amount extracting unit among a plurality of feature amount extracting units that extract feature amounts. That is, in the second embodiment, selection of a first feature quantity extraction unit 79a or a second feature quantity extraction unit 79b described later is accepted. It should be noted that the selection of either one of the first feature quantity extraction means 79a or the second feature quantity extraction means 79b may be accepted, or both selections may be accepted. It goes without saying that selection may be accepted from a plurality of feature quantity extraction means including other feature quantity extraction means.

  The feature quantity extraction unit 79 includes a first feature quantity extraction unit 79a and a second feature quantity extraction unit 79b. The first feature amount extraction unit 79a extracts a pixel value related to a color component as a feature amount for each pixel of the acquired multi-value image or the generated distance image. The second feature amount extraction unit 79b extracts edge strengths in two different directions as feature amounts for each pixel of the acquired multi-valued image or generated distance image. When the feature amount is a pixel value related to a color component, a difference in color can be detected, and when the feature amount is an edge intensity in two different directions, the presence of a defect in the vicinity of the contour line can be detected.

  The distribution range calculation unit 73 calculates a distribution range for performing non-defective product determination based on the feature amount extracted by the feature amount extraction unit 79 that has received the selection. When the selection of the first feature quantity extraction means 79a is accepted, the distribution range calculation means 73a accepts the selection of the second feature quantity extraction means 79b. Calculate the range. Information on the calculated distribution range is stored in the distribution range information storage unit 232 of the storage unit 23 as a function expression indicating the boundary of the distribution range, coordinate values, threshold values, and the like.

  The multi-value image acquisition unit 74 acquires a multi-value image (second multi-value image) related to the determination target imaged by the camera 1. The target distance image generation means 75 performs the phase analysis of the fringe pattern based on the fringe pattern image obtained by capturing the fringe pattern projected by irradiating light from the light 9 with the camera 1, thereby A target distance image (second distance image) specifying a gray value for each pixel according to the distance from the surface to the camera 1 is generated. The method for generating the target distance image is the same as that for the non-defective distance image generating means 72.

  The position adjustment unit 76 aligns the multi-value image related to the determination target and the multi-value image related to the non-defective product, and aligns the target distance image related to the generated determination target and the non-defective product distance image related to the non-defective product. Specifically, an average image of a plurality of multi-value images or non-defective product distance images related to non-defective products is calculated, and alignment with the average image is performed. The means for performing alignment between multi-value images or distance images is not particularly limited as long as it is a known technique. For example, the positions of the feature points of both multi-valued images or distance images may be detected by pattern matching, and alignment may be performed using an upper limit value and a lower limit value that can be determined to match. In addition, the degree of coincidence between both multi-valued images or distance images may be calculated by calculating a normalized correlation or the like, and alignment may be performed so that the degree of coincidence is greater than a predetermined value. Of course, the alignment may be performed so that the contour lines, areas, centroids, and the like of both the multi-valued images or distance images match.

  The selected feature amount extraction unit 79c is the same feature amount extraction unit as the feature amount extraction unit 79 that has received selection by the selection reception unit 78 for each pixel of the acquired multi-value image or the generated target distance image related to the determination object. Extract features. For example, when the selection of the first feature quantity extraction unit 79a is accepted, the pixel value related to the color component is extracted as the feature quantity for each pixel of the multi-value image or the target distance image related to the determination target, and the second feature quantity When the selection of the extraction unit 79b is received, edge strengths in two different directions are extracted as feature amounts for each pixel of the multi-value image or the target distance image related to the determination target.

  The determination unit 77 extracts the corresponding feature amount by the feature amount extraction unit 79 that has received the selection for each pixel of the acquired multi-value image or the generated target distance image, and the extracted feature amount is calculated. It is determined whether or not the multi-valued image or the non-defective distance image regarding the non-defective product is included in the same feature amount distribution range. If it is determined that all the extracted feature values are included in the distribution range, it is determined that the determination target object is a non-defective product. If it is determined that none of the extracted feature values is included, the determination target object is a non-defective product. It is determined that it is not.

  FIG. 7 is a flowchart showing a non-defective product information collection process performed by the main control unit 21 of the image processing apparatus 2 according to the second embodiment of the present invention. The main control unit 21 of the image processing apparatus 2 acquires a plurality of multi-value images (first multi-value images) related to non-defective products, which are captured by the camera 1 (step S701). The main control unit 21 stores the acquired multi-value image data of the multi-value image group in the non-defective image data storage unit 231 of the storage unit 23 (step S702).

  The main control unit 21 acquires a fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1 for a plurality of non-defective products (step S703), and executes phase analysis of the fringe pattern. As a result, a plurality of non-defective product distance images specifying the gray value for each pixel in accordance with the distance from the non-defective product surface to the camera 1 are generated (step S704). The main control unit 21 stores the generated distance image data of the non-defective distance image group in the storage unit 23 (step S705).

  The main control unit 21 receives the selection of the feature amount extraction unit 79 that extracts the feature amount from the multi-valued image or the good product distance image (step S706). The feature quantity extraction unit 79 to be selected is a first feature quantity extraction unit 79a that extracts a pixel value related to a color component as a feature quantity for each pixel of the acquired multi-valued image or the generated non-defective distance image, and acquired. The second feature amount extraction unit 79b extracts edge strengths in two different directions as feature amounts for each pixel of the multi-valued image or the generated non-defective distance image. Note that the feature amount extraction unit 79 that accepts selection may be one or plural.

  The main control unit 21 extracts the feature amount for each pixel of the multi-valued image or the non-defective distance image by the feature amount extracting unit 79 that has received the selection (step S707), and the feature from which the distribution range for performing the non-defective product determination is extracted. Calculation is performed for each amount (steps S708a and S708b). When there is only one feature amount extraction unit 79 that has received selection, for example, only by step S708a, and when there are two feature amount extraction units 79 that have received selection, steps S708a and S708b respectively. A corresponding distribution range is calculated for each extracted feature amount. The main control unit 21 stores information on the calculated distribution range in the distribution range information storage unit 232 of the storage unit 23 as a function expression indicating the boundary of the distribution range, coordinate values, threshold values, and the like (step S709).

  FIG. 8 is a flowchart showing a non-defective product determination process performed by the main control unit 21 of the image processing apparatus 2 according to the second embodiment of the present invention. The main control unit 21 of the image processing apparatus 2 acquires a multi-value image (second multi-value image) related to the determination target imaged by the camera 1 (step S801). The main control unit 21 aligns the acquired multi-value image related to the determination target object and the multi-value image related to the non-defective product (first multi-value image) (step S802). Specifically, an average image of a plurality of multi-valued images related to non-defective products is calculated, and alignment with the average image is performed.

  The main control unit 21 obtains a fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1 (step S803), and performs a phase analysis of the fringe pattern to thereby determine a determination target. A target distance image specifying a gray value for each pixel according to the distance from the surface of the object to the camera 1 is generated (step S804).

  The main control unit 21 performs alignment between the target distance image related to the generated determination target object and the non-defective distance image related to the non-defective product (step S805). Specifically, an average distance image of a plurality of non-defective product distance images related to the non-defective product is calculated, and alignment is performed with the average distance image.

  The main control unit 21 extracts a feature amount by the feature amount extraction unit 79 that has received the selection for each pixel of the acquired multi-value image or the generated target distance image regarding the determination object (step S806). For example, when the selection of the first feature quantity extraction unit 79a that extracts the pixel value related to the color component is accepted as the feature quantity extraction unit 79, the color of each pixel of the multi-value image or the target distance image related to the determination target is determined. When a pixel value related to a component is extracted as a feature quantity and selection of the second feature quantity extraction means 79b that extracts edge strengths in two different directions is accepted, a pixel of a multi-value image or target distance image related to the determination target Each time, edge strengths in two different directions are extracted as feature amounts.

  Whether the main control unit 21 includes the feature value extracted for each pixel of the multi-value image or the target distance image in the distribution range corresponding to the same feature value of the calculated multi-value image or the good product distance image. It is determined whether or not (step S807). When the main control unit 21 determines that all the feature values of the extracted multi-value image or target distance image are included in the distribution range (step S807: YES), the main control unit 21 determines that the determination target is a non-defective product. It is determined that there is, and information indicating that the determination target is a non-defective product is displayed on the display screen of the display device 3 as a determination result (step S808).

  When the main control unit 21 determines that even one feature amount of the extracted multi-value image or target distance image is not included in the distribution range (step S807: NO), the main control unit 21 determines that the determination target object is It is determined that the product is not a non-defective product, and information indicating that the determination target is not a good product is displayed on the display screen of the display device 3 as a determination result (step S809). When a plurality of selections by the feature amount extraction unit 79 are received, one of the determination results of the non-defective product determination using the feature amount extracted by the feature amount extraction unit 79 that has received the selection is not a non-defective product. When it is determined, it may be determined that the determination target is not a good product.

  Further, in step S806 described above, the feature amount is extracted by the feature amount extraction unit 79 that has received the selection for each pixel of the acquired multi-value image related to the determination target object or the generated target distance image. With respect to both the multi-value image relating to the determination target object and the generated target distance image, the feature amount can be extracted by the feature amount extraction unit 79 that accepts selection for each pixel. That is, for each pixel of the acquired multi-valued image relating to the determination object, the feature amount is extracted by the feature amount extraction unit 79 that has received the selection, and the feature amount extraction that has received the selection for each pixel of the generated target distance image It is also possible to extract feature quantities by means 79 (feature quantities may be the same or different). Then, for each of the acquired multi-value image related to the determination object and the generated target distance image, it is determined whether or not the feature amount is included in the distribution range, and attention is paid to the acquired multi-value image related to the determination object. The first non-defective product determination and the second non-defective product determination when paying attention to the generated target distance image are performed. For example, if it is determined that the feature amount is included in the distribution range in any non-defective product determination, it is finally determined that the feature amount is in the distribution range, and the feature amount is included in the distribution range in any one non-defective product determination. If it is determined that it is not, it is finally determined that the product is defective. As described above, the distribution range calculation unit 73 may calculate not only the generated distance image but also the distribution range for performing the non-defective product determination of the gray value for each pixel of the multi-value image. It may be determined whether or not each gray value is included in the gray value distribution range for each pixel of the non-defective product multi-valued image regarding the calculated good product. An accepting unit (such as a mouse or a monitor) that accepts user settings for conditions for finally determining that the product is non-defective may be provided.

  Here, when the selection of the first feature quantity extraction unit 79a is accepted, that is, when the feature quantity to be extracted is a pixel value related to a color component, for each pixel of a multi-value image related to a good product or a good product distance image. An average value of pixel values related to color components is calculated, and a distribution range of pixel values is calculated for each pixel of a multi-value image related to good products or a good product distance image based on the calculated average values of pixel values related to color components. FIG. 9 is a functional block diagram of the distribution range calculation unit 73a when the selection of the first feature amount extraction unit 79a of the image processing apparatus 2 according to Embodiment 2 of the present invention is accepted.

  As shown in FIG. 9, the distribution range calculation unit 73 a includes a pixel average value calculation unit 731 and a pixel value distribution range calculation unit 732. The main control unit 21 of the image processing unit 7 controls processing operations of the pixel average value calculating unit 731 and the pixel value distribution range calculating unit 732.

  The pixel average value calculating unit 731 calculates a pixel value related to a color component, for example, an average value of gray values, for each pixel of the acquired multi-value image related to the good product or the generated good product distance image. The pixel value distribution range calculation means 732 is a multidimensional space in which each color component is a coordinate axis for each pixel of a multi-valued image or a non-defective product distance image related to a non-defective product based on a pixel value related to the calculated color component, for example, an average value of gray values. The distribution range of pixel values at is calculated.

  For example, assume a three-dimensional coordinate system having R, G, and B components as coordinate axes. In the method of simply calculating the average value of the pixel values related to the R component, G component, and B component, and calculating the distribution range of the pixel values of each component to perform the non-defective product determination, the correct determination is made based on the color components. There are cases where it is not possible.

  For example, when the non-defective product is distributed in a color including a color similar to white to dark gray, that is, the R component pixel values are 30 to 240, the G component pixel values are 30 to 240, and the B component pixel values are 30 to 240. A determination object having a color close to red, such as a determination object having an R component pixel value of 200, a G component pixel value of 50, and a B component pixel value of 100, is clearly not a good product. However, it may be determined that the pixel value is included in the distribution range of the pixel values for each component. Therefore, R component, G component, and B component are not individually determined, but depending on whether they are included in the pixel value distribution range in the virtual three-dimensional space with R component, G component, and B component as coordinate axes. Judge whether it is a good product.

  FIG. 10 is an exemplary diagram showing a distribution range of pixel values when selection by the first feature quantity extraction unit 79a of the image processing apparatus 2 according to Embodiment 2 of the present invention is accepted. In FIG. 10, on the basis of a multi-value image or a non-defective distance image regarding a plurality of non-defective products, points 60, 60,... In a three-dimensional space having pixel values of each color component as R, G, and B components orthogonal to each other. Plotted as. Then, an average value Av of pixel values of each color component is calculated, and a distribution range 61 for performing non-defective product determination in a three-dimensional space is calculated around the calculated average value Av.

  For example, at points 62 and 63 where pixel values relating to each color component of each pixel of the multi-value image or the target distance image of the determination target are plotted, the point 63 is farther from the average value Av. However, the point 63 is within the distribution range 61, whereas the point 62 is plotted outside the distribution range 61. Therefore, the determination target object plotted as the point 63 is determined to be non-defective, and the determination target object plotted as the point 62 is determined to be non-defective.

  The calculation method of the distribution range 61 is not particularly limited. For example, a Euclidean distance that is a spatial distance may be used, or a Mahalanobis distance may be used. Hereinafter, a case where the Mahalanobis distance is used will be described.

First, the Mahalanobis distance D _M of the pixel value related to the color component for each pixel is the pixel vector x (r, g, The average value of b) is the average value vector μ _i (r bar, g bar, b bar), and the inverse matrix of the covariance matrix is Σ _i ^−1. In this case, it can be calculated as (Equation 1) using a vector determinant.

In (Expression 1), λ _j is an eigenvalue, and the vector φ _j is an eigenvector corresponding to the eigenvalue λ _j . That is, the Mahalanobis distance D _M calculated by (Equation 1) is obtained by decomposing the distance between the point x (r, g, b) and the center of gravity (average value) of the distribution range 61 into eigenvector φ _j direction components, and eigenvector φ _j It can be said that the direction component is normalized by the variance λ _j . Note that i in (Equation 1) indicates the distinction of the distribution, and the above calculation does not have a special meaning because the distribution range 61 is one.

Since it is a virtual three-dimensional space, in (Expression 1), n = 3, there are three eigenvalues λ _j and eigenvectors φ _j , and the eigenvector φ _j is also a three-dimensional matrix. That is, the variance-covariance matrix Σ _i can be expressed as (Equation 2).

  In (Expression 2), V represents the dispersion, and the dispersion of each of the R component, the G component, and the B component can be obtained by (Expression 3).

  In (Expression 2), Cov represents covariance, and the covariance between the R component and the G component, the covariance between the R component and the B component, and the covariance between the G component and the B component are ( It can be obtained from equation 4).

  (Formula 5) is obtained by multiplying n by the denominator and numerator of (Formula 4).

From (Equation 5), based on the multi-value image or the non-defective distance image regarding the non-defective product, the sum (average value) of the pixel values regarding the color component, the square sum of the pixel values regarding the color component, and two sets of colors for each pixel. It can be seen that the variance-covariance matrix Σ _i can be obtained by calculating the sum of the values obtained by multiplying the component pixel values. The sum total of pixel values related to color components (average value), the sum of squares of pixel values related to color components, and the sum of values obtained by multiplying pixel values of two sets of color components are stored in the non-defective image data storage unit 231. The distribution range 61 for determining that the product is non-defective is calculated.

Next, regarding the pixel value related to the color component for each pixel of the multivalued image or the distance image related to the determination object, the Mahalanobis distance D _M calculated by (Equation 1), that is, from the center of gravity of the distribution range 61 to the pixel value related to the color component. Is normalized by the distance between the boundary of the distribution range 61 and the center of gravity in the direction from the center of gravity of the distribution range 61 to the pixel value relating to the color component, and the pixel value is greater than the predetermined value. Not included. Therefore, since the color has changed beyond the non-defective range, it can be determined that the color is not good.

  As another determination method, for example, a distance from the center of gravity of a range of a shape like a virtual ellipsoid (hereinafter referred to as an elliptical sphere range) to a pixel value for each pixel of the multivalued image or target distance image related to the determination target. The Euclidean distance is calculated, and the calculated Euclidean distance is the distance between the center of gravity of the elliptic sphere range and the center of gravity in the direction from the centroid of the elliptic sphere range to the pixel value for each pixel of the multi-valued image or target distance image regarding the determination target. It is determined whether or not it is included in the cross-correlation distribution range based on whether or not it is smaller than a predetermined value. By using the Euclidean distance, it is possible to calculate the degree of deviating from the cross-correlation distribution range determined as non-defective without the sensitivity greatly changing due to the size of the distribution spread.

  Further, it is not necessary to limit the distribution range 61 to an elliptical sphere range, and a rectangular parallelepiped-shaped range inscribed in the distribution range 61 may be calculated virtually.

  FIG. 11 is a flowchart showing the procedure of the distribution range calculation process when the main control unit 21 of the image processing apparatus 2 according to Embodiment 2 of the present invention receives selection of the first feature quantity extraction unit 79a. is there. Such processing corresponds to step S708a in FIG.

In FIG. 11, the main control unit 21 of the image processing apparatus 2 calculates based on a plurality of multi-value images or non-defective distance images regarding non-defective products and stores them in the non-defective image data storage unit 231. Read out the sum of pixel values related to the color components for each pixel of the distance image (average value), the sum of squares of the pixel values related to the color components, and the sum of the values multiplied by the pixel values related to two sets of color components (step S1101), A variance / covariance matrix Σ _i of pixel values relating to color components for each pixel is created (step S1102). That is, the variance-covariance matrix Σ _i shown in (Expression 2) is created.

The main control unit 21 calculates the eigenvalue λ _j and eigenvector φ _j of the variance-covariance matrix Σ _i (step S1103), and stores them in the distribution range information storage unit 232 as distribution range information related to the distribution range (step S1104).

The determination as to whether or not the image is included in the distribution range in step S807 in FIG. 8 is performed by using a pixel value related to a color component as a feature value for each pixel of the acquired multi-value image or the generated target distance image. When the Mahalanobis distance D _M that can be calculated and calculated by (Equation 1) is larger than a predetermined value, it is determined that it is not included in the distribution range 61, and it can be determined that it is not a non-defective product.

  Next, when the selection of the second feature quantity extraction unit 79b is accepted, that is, when the feature quantity to be extracted is the edge strength in two different directions, the multi-value image or the distance image for each non-defective product is obtained for each pixel. An average value of edge intensities in two different directions is calculated, and a cross-correlation distribution range of edge intensities in two different directions is calculated for each pixel of the multi-valued image or non-defective product distance image regarding the non-defective product with the calculated average value as the center. FIG. 12 is a functional block diagram of the distribution range calculation unit 73b when the selection of the second feature amount extraction unit 79b of the image processing apparatus 2 according to Embodiment 2 of the present invention is accepted.

  As shown in FIG. 12, the distribution range calculation unit 73b includes a direction-specific edge strength calculation unit 734, an edge strength average value calculation unit 735, and an edge strength distribution range calculation unit 736. The main control unit 21 of the image processing unit 7 controls processing operations of the direction-specific edge intensity calculating unit 734, the edge intensity average value calculating unit 735, and the edge intensity distribution range calculating unit 736.

  The direction-specific edge strength calculation means 734 calculates edge strengths in two different directions for each pixel of the acquired multi-value image or the generated good product distance image. Although the direction in which the edge strength is calculated is not particularly limited, in the second embodiment, the edge strength is calculated for each of the row direction and the column direction of the two-dimensional image which are two different directions. Specifically, the edge intensity in the row direction and the column direction may be calculated using a 3 × 3 window template used in the Sobel filter.

  The edge strength average value calculation means 735 calculates an average value of edge strengths in two different directions for each pixel of the multi-valued image or the non-defective distance image regarding the non-defective product based on the calculated edge strengths in two different directions for each pixel. .

  The edge intensity distribution range calculation means 736 calculates the cross-correlation distribution range of the edge intensity in two different directions for each pixel of the multi-valued image or the non-defective distance image regarding the non-defective product with the calculated average value as the center.

  FIG. 13 is an exemplary diagram of a cross-correlation distribution range of edge strengths in two different directions for each predetermined pixel of the image processing apparatus 2 according to Embodiment 2 of the present invention. In FIG. 13, the edge strength ex in the row direction in the two-dimensional image is taken as the X axis, and the edge strength ey in the column direction is taken as the Y axis.

The edge strength average value calculation means 735 calculates the average value of the edge strength ex in the row direction and the average value of the edge strength ey in the column direction in the two-dimensional image for each pixel of the multi-value image or the good product distance image related to the good product. Therefore, the average value vector E _AVE of the edge strength for each pixel of the multi-value image or the non-defective distance image regarding the non-defective product is located on a straight line having the edge angle θ as the inclination angle. For example, this is a case where there is a large variation in edge intensity for each pixel of the acquired multi-value image (first multi-value image) related to a plurality of non-defective products or the generated good-quality distance image (first distance image) related to a plurality of non-defective products. When the edge directions are almost aligned, the average value vector E _AVE of the edge strengths of the multi-valued image or the good-quality distance image for the non-defective product is mutually displayed as a virtual elliptical area centered on the point where the major axis and the minor axis intersect. A correlation distribution region 91 is calculated. Therefore, the edge strength for each pixel of the multi-value image or the target distance image related to the determination target is calculated, and the edge strength ex in the row direction and the edge strength ey in the column direction in the two-dimensional image are plotted for the same pixel. By determining whether or not the point 92 is included in the cross-correlation distribution range 91, it can be determined whether or not the determination target is a non-defective product.

  The calculation method of the cross-correlation distribution range 91 is not particularly limited. For example, a Euclidean distance that is a spatial distance may be used, or a Mahalanobis distance may be used. Hereinafter, a case where the Mahalanobis distance is used will be described.

  In this case, the distance calculation means 70 of the determination means 77 shown in FIG. 6 determines the object to be determined from the centroid of the distribution range, for example, the centroid (center point) of the cross-correlation distribution range 91 which is a virtual elliptical area in the example of FIG. Normalize the distance to the edge point (feature amount) for each pixel of the multi-valued image or target distance image with respect to the boundary and the centroid of the cross-correlation distribution range 91 in the direction from the centroid to the edge point (feature amount) Calculate the Mahalanobis distance.

Mahalanobis distance D _M of edge intensity for each pixel
Is two different directions, i.e. the average value of each average value vector μ _i (e _x bar in the example of FIG. 13 X-axis direction and Y each pixel in the direction of the edge intensity vector _{x (e x, e y)} , e _y bar), and the inverse of the variance-covariance matrix Σ _i ^-1 In this case, it can be calculated as (Expression 6) using a vector determinant.

In (Expression 6), λ _j is an eigenvalue, and vector φ _j is an eigenvector corresponding to the eigenvalue λ _j . That Mahalanobis distance D _M is calculated by (Equation 6), the point x (e _x, e _y) the distance between the center of gravity of the cross-correlation distribution range 91 (average value), and degraded eigenvector phi _j direction component, eigenvector It can be said that the φ _j direction component is normalized by the variance λ _j . Note that i in (Equation 6) indicates the distinction of the distribution, and since the cross correlation distribution range 91 is one in the above calculation, it does not have any special meaning.

Unlike FIG. 10, since it is a two-dimensional space in two different directions, in (Equation 6), n = 2, eigenvalue λ _j and eigenvector φ _j are two, and eigenvector φ _j is also two-dimensional. It becomes the matrix of. That is, the variance-covariance matrix Σ _i can be expressed as (Equation 7).

  In (Expression 7), V represents the dispersion, and the dispersion of the edge strength in the X-axis direction and the edge strength in the Y-axis direction can be obtained by (Expression 8).

  In (Expression 7), Cov represents covariance, and the covariance between the edge strength in the X-axis direction and the edge strength in the Y-axis direction can be obtained by (Expression 9).

From (Equation 9), based on a multi-value image or a non-defective distance image regarding a non-defective product, for each pixel, the sum (average value) of edge strengths in two different directions, the square sum of edge strengths in two different directions, and two different directions It can be seen that the variance-covariance matrix Σ _i can be obtained by calculating the total sum of the values obtained by multiplying the edge intensities at. The sum (average value) of edge strengths in two different directions, the square sum of edge strengths in two different directions, and the sum of values obtained by multiplying the edge strengths in two different directions are stored in the non-defective image data storage unit 231. The cross-correlation distribution range 91 for determining that the product is non-defective is calculated.

Next, regarding the edge strength in two different directions (X-axis direction and Y-axis direction) for each pixel of the multivalued image or the target distance image related to the determination target, the Mahalanobis distance D _M calculated by (Expression 6), that is, mutual The distance from the centroid of the correlation distribution range 91 to the edge point (feature) is normalized by the distance between the boundary of the cross-correlation distribution range 91 and the centroid in the direction from the centroid of the cross-correlation distribution range 91 to the edge point (feature). If the converted value is larger than the predetermined value, the edge point is not included in the cross-correlation distribution range 91. Therefore, since the edge angle indicates that it has fluctuated beyond the non-defective range, it can be determined that the shape of the contour has changed greatly, and it can be determined that the edge is not non-defective.

  FIG. 14 is a flowchart showing the procedure of the distribution range calculation process when the main control unit 21 of the image processing apparatus 2 according to Embodiment 2 of the present invention accepts the selection of the second feature quantity extraction unit 79b. is there. Such processing corresponds to step S708b in FIG.

In FIG. 14, the main control unit 21 of the image processing apparatus 2 calculates and stores a multi-value image or a good product distance image based on a multi-value image or a good product distance image related to a good product in two different directions for each pixel. The sum of the edge strengths (average value), the sum of squares of the edge strengths in two different directions, and the sum of values obtained by multiplying the edge strengths in the two different directions are read (step S1401), and the edge strengths in the two different directions for each pixel are read. A variance-covariance matrix Σ _i is created (step S1402).

The main control unit 21 calculates the eigenvalue λ _j and eigenvector φ _j of the variance-covariance matrix Σ _i (step S1403), and stores them in the distribution range information storage unit 232 as distribution range information related to the cross-correlation distribution range (step S1404). .

Then, the determination as to whether or not the image is included in the distribution range in step S807 in FIG. 8 is based on the edge strength in two different directions as the feature value for each pixel of the obtained multi-value image or the generated target distance image. When the Mahalanobis distance D _M that can be calculated by (Equation 6) is larger than a predetermined value, it is determined that it is not included in the cross-correlation distribution range 91, and it is determined that it is not a good product. it can.

  FIG. 15 is a flowchart showing a non-defective product determination process when the main control unit 21 of the image processing apparatus 2 according to the second embodiment of the present invention receives selection of the second feature quantity extraction unit 79b. . The main control unit 21 of the image processing apparatus 2 acquires a multi-value image related to the determination target imaged by the camera 1 (step S1501). The main control unit 21 aligns the acquired multi-value image related to the determination target object and the multi-value image related to the non-defective product (step S1502). Specifically, an average image of a plurality of multi-valued images related to non-defective products is calculated, and alignment with the average image is performed. The means for aligning the multi-value images is not particularly limited as long as it is a well-known technique.

  The main control unit 21 obtains a fringe pattern image obtained by imaging the fringe pattern projected by irradiating light from the light 9 with the camera 1 (step S1503), and executes a phase analysis of the fringe pattern to determine a determination target. A target distance image specifying a gray value for each pixel according to the distance from the surface of the object to the camera 1 is generated (step S1504).

  The main control unit 21 performs alignment between the target distance image related to the generated determination target object and the non-defective distance image related to the non-defective product (step S1505). Specifically, an average distance image of a plurality of non-defective product distance images related to the non-defective product is calculated, and alignment is performed with the average distance image.

  The main control unit 21 calculates edge strengths in two different directions for each pixel of the acquired multi-value image or the generated target distance image related to the determination object (step S1506). The direction in which the edge strength is calculated is not particularly limited, but the edge strength is calculated for each of the column direction and the row direction in the two-dimensional image which are two different directions, as in the method described above.

  The main control unit 21 determines that the calculated edge intensity for each pixel of the multi-value image or the target distance image related to the determination target object is within the cross-correlation distribution range of the edge strength for each pixel of the calculated multi-value image or the good distance image. It is determined whether it is included (step S1507). When the main control unit 21 determines that all of the calculated edge intensities are included in the cross-correlation distribution range (step S1507: YES), the main control unit 21 determines that the determination target is a non-defective product and determines Information indicating that the object is a non-defective product is displayed on the display screen of the display device 3 as a determination result (step S1508). When the main control unit 21 determines that even one calculated edge strength is not included in the cross-correlation distribution range (step S1507: NO), the main control unit 21 determines that the determination target is not a non-defective product. Then, information indicating that the determination target is not a good product is displayed on the display screen of the display device 3 as a determination result (step S1509).

  In the second embodiment described above, the case where the cross-correlation distribution range is a virtual elliptical area has been described. However, the present invention is not particularly limited to the elliptical area.

  As described above, according to the second embodiment, the multi-value image (second multi-value image) or the target distance image (second distance image) related to the acquired determination target object is a non-defective product. Since at least one means for extracting a feature quantity as a basis for determining whether or not a defect has occurred in the vicinity of the contour line by selecting an appropriate feature quantity extraction means 79 can be selected. Even if it exists, the presence of a defect can be detected, and a difference in color can be detected with high sensitivity. For example, when the feature amount extraction unit 79 that has received a selection extracts a pixel value related to a color component as a feature amount, it is possible to detect the presence or absence of shading variation due to a color difference, and the feature amount extraction unit 79 that has received a selection. When edge strengths in two different directions are extracted as feature amounts, it is possible to detect the presence of a defect near the contour line.

  The present invention is not limited to the above-described embodiments, and various changes and improvements can be made within the scope of the present invention. Two different directions are used as the row direction and the column direction of the two-dimensional image. However, the two directions are not particularly limited, and the two directions are orthogonal to each other. The color components are not limited to the R component, the G component, and the B component. , A cyan component, a magenta component, and a yellow component.

1 Camera (imaging means)
2 Image processing device 3 Display device (image display means)
4 portable recording medium 5 computer program 6 external control device 7 image processing unit 8 image display unit 9 light (light irradiation means)
21 Main Control Unit 22 Memory 23 Storage Unit 24 Input Unit 25 Output Unit 26 Communication Unit 27 Auxiliary Storage Unit 28 Internal Bus

Claims (11)

In an image processing apparatus that performs a non-defective product determination by comparing a multi-value image obtained by imaging a determination target with a multi-value image group related to a good product,
A non-defective image acquisition means for acquiring a plurality of first multi-value images related to the non-defective product imaged by the imaging means;
A non-defective distance image generating means for generating a first distance image obtained by converting the distance from the imaging means into a gray value for each pixel of the acquired first multi-valued image;
For each pixel of the first multi-valued image, for each pixel of the first multi-valued image or the first distance image, a first feature amount extracting means for extracting a pixel value related to a color component as a feature amount. Selection accepting means for accepting selection of at least one feature quantity extracting means among a plurality of feature quantity extracting means including at least second feature quantity extracting means for extracting edge strengths in two different directions as feature quantities;
The gray value of each at least one pixel of the acquired first multivalued image and generated the first distance image, the distribution range calculation means for calculating a distribution range for performing good judgment,
Multi-valued image acquisition means for acquiring a second multi-valued image related to the determination target;
Target distance image generation means for generating a second distance image obtained by converting the distance from the imaging means into a gray value for each pixel of the acquired second multi-valued image;
Determination means for determining whether or not the gray value for each pixel of at least one pixel of the acquired second multi-valued image and the generated second distance image is included in the corresponding distribution range, and
The distribution range calculation means displays the first distribution range when the selection of the first feature quantity extraction means is accepted, and the second distribution when the selection of the second feature quantity extraction means is accepted. The range is calculated based on the feature amount extracted by the feature amount extraction unit that has accepted the selection,
The determination unit extracts a feature amount by a feature amount extraction unit that has received a selection for each pixel of the second multi-valued image or the second distance image, and the first corresponding to the extracted feature amount An image processing apparatus is characterized in that it is determined whether or not the distribution range is included in the distribution range or the second distribution range.   The non-defective product distance image generating unit or the target distance image generating unit includes a plurality of the first multi-value images or the second multi-value images captured by the plurality of image capturing units. 2. The image processing apparatus according to claim 1, wherein the first distance image or the second distance image is generated based on the first distance image.   The imaging unit and a light irradiation unit that irradiates light to the determination target, wherein the non-defective distance image generation unit or the target distance image generation unit is the first multi-value imaged by the imaging unit. The image processing apparatus according to claim 1, wherein the first distance image or the second distance image is generated based on the image or the second multi-valued image and the irradiated light. . When receiving the selection of the first feature amount extraction unit, the distribution range calculation unit,
Pixel average value calculating means for calculating an average value of pixel values related to color components for each pixel of the first multi-valued image;
Pixels for calculating the first distribution range of pixel values in a multidimensional space with each color component as a coordinate axis for each pixel of the first multi-valued image, based on the average value of the pixel values relating to the calculated color component A value distribution range calculation means, and
The determination means calculates a pixel value related to a color component for each pixel of the second multi-value image, and is included in the first distribution range of pixel values in a multi-dimensional space with each color component as a coordinate axis. To judge whether or not
When receiving the selection of the second feature amount extraction means, the distribution range calculation means,
Direction-specific edge strength calculating means for calculating edge strength in two different directions for each pixel of the first multi-valued image or the first distance image;
Based on the calculated edge strength, for each pixel of the first multi-value image or the first distance image, an edge strength average value calculating means for calculating an average value of edge strengths in two different directions;
Edge strength for calculating the cross-correlation distribution range of edge strengths in two different directions as the second distribution range for each pixel of the first multi-valued image or the first distance image with the calculated average value as the center A distribution range calculation means, and
The determining means calculates edge strengths in two different directions for each pixel of the second multi-valued image or the second distance image, and the calculated edge strengths are included in the corresponding cross-correlation distribution range. 4. The image processing apparatus according to claim 1, wherein the image processing apparatus is configured to determine whether or not the image processing apparatus is present.   5. The image processing apparatus according to claim 4, wherein the direction-specific edge strength calculating means calculates the edge strength in two directions orthogonal to each other.   6. The image processing apparatus according to claim 5, wherein the direction-specific edge strength calculating means calculates the edge strength in two directions of a row direction and a column direction in a two-dimensional image.   7. The edge intensity distribution range calculation means calculates the cross-correlation distribution range as the second distribution range as a virtual ellipse region. The image processing apparatus described.   A position adjustment unit is provided for aligning the first multi-value image and the second multi-value image or aligning the first distance image and the second distance image. Item 8. The image processing apparatus according to any one of Items 1 to 7.   In the case where the selection accepting unit accepts the selection of a plurality of feature amount extracting units, any one of the feature amounts extracted by the feature amount extracting unit that has received the selection by the determining unit is When it is determined that the object is not included in the first distribution range or the second distribution range corresponding to each feature amount, the determination object is determined not to be non-defective. The image processing apparatus according to claim 1. In an image processing method that can be executed by an image processing apparatus that performs a non-defective product determination by comparing a multi-value image obtained by imaging a determination target with a multi-value image group related to a non-defective product,
Acquiring a plurality of first multi-valued images related to non-defective products imaged by the imaging means;
Generating a first distance image obtained by converting the distance from the imaging unit into a gray value for each pixel of the acquired first multi-valued image;
For each pixel of the first multi-valued image, for each pixel of the first multi-valued image or the first distance image, a first feature amount extracting means for extracting a pixel value related to a color component as a feature amount. Receiving a selection of at least one feature quantity extraction means among a plurality of feature quantity extraction means including at least a second feature quantity extraction means for extracting edge strengths in two different directions as feature quantities;
Calculating the gray value of each at least one pixel of the acquired first multivalued image and generated the first distance image, the distribution range for performing good judgment,
Obtaining a second multi-valued image relating to the determination object;
Generating a second distance image obtained by converting the distance from the imaging unit into a gray value for each pixel of the acquired second multi-valued image;
Determining whether a gray value for each pixel of at least one pixel of the acquired second multi-valued image and the generated second distance image is included in the corresponding distribution range, and
When the selection of the first feature quantity extraction unit is accepted, the selection is accepted, and when the selection of the second feature quantity extraction unit is accepted, the second distribution range is accepted. Calculated based on the feature amount extracted by the feature amount extraction means,
For each pixel of the second multi-value image or the second distance image, a feature amount is extracted by a feature amount extraction unit that has received a selection, and the first distribution range corresponding to the extracted feature amount or the An image processing method comprising determining whether or not the second distribution range is included. In a computer program that can be executed by an image processing apparatus that performs a non-defective product determination by comparing a multi-value image obtained by imaging a determination target with a multi-value image group related to a non-defective product,
The image processing apparatus;
Non-defective image acquisition means for acquiring a plurality of first multi-value images related to non-defective products imaged by the imaging means,
Non-defective distance image generation means for generating a first distance image obtained by converting the distance from the imaging means into a gray value for each pixel of the acquired first multi-valued image,
For each pixel of the first multi-valued image, for each pixel of the first multi-valued image or the first distance image, a first feature amount extracting means for extracting a pixel value related to a color component as a feature amount. Selection accepting means for accepting selection of at least one feature quantity extracting means among a plurality of feature quantity extracting means including at least second feature quantity extracting means for extracting edge strengths in two different directions as feature quantities;
Acquired first gray value of at least every one of the pixels of the multivalued image and the generated the first distance image, the distribution range calculation means for calculating a distribution range for performing good judgment,
Multi-valued image obtaining means for obtaining a second multi-valued image relating to the determination object;
Target distance image generation means for generating a second distance image obtained by converting the distance from the imaging means into a gray value for each pixel of the acquired second multi-valued image; and
Function as a determination means for determining whether or not the gray value for each pixel of at least one pixel of the acquired second multi-valued image and the generated second distance image is included in the corresponding distribution range;
When the selection of the first feature amount extraction unit is accepted, the distribution range calculation unit displays the first distribution range, and when the selection of the second feature amount extraction unit is accepted, the second distribution The range functions as a means for calculating based on the feature amount extracted by the feature amount extraction means that accepts the selection,
The determination unit extracts the feature amount by the feature amount extraction unit that has received selection for each pixel of the second multi-valued image or the second distance image, and the first feature corresponding to the extracted feature amount A computer program that functions as means for determining whether or not the distribution range is included in the second distribution range.

Images