An interest area fragmentation industrial vision detection method and system

Abstract

This invention relates to an industrial visual inspection method for fragmented regions of interest (ROIs), characterized by the following steps: Step 1, offline construction step, obtaining the transformation matrix M1 between the initial coordinate system and the workpiece coordinate system, and the transformation matrix set M2 between the workpiece coordinate system and the offline ROI set R1; Step 2, online construction step, based on the primary coordinate system and the transformation matrix M1, inversely mapping the offline secondary positioning box P1 to obtain the real-time secondary positioning box P2' of the workpiece under test; using P2' to establish a secondary coordinate system, and using the secondary coordinate system and the transformation matrix set M2, inversely mapping the offline ROI set R1 to obtain the real-time ROI set R2' of the workpiece under test; Step 3, cropping the real-time scene grayscale image according to the real-time ROI set R2'. Compared with the prior art, this invention can quickly calculate the real-time ROI set R2', and performs independent positioning after each acquisition of the real-time scene grayscale image, is less affected by other factors, and has high positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of industrial visual inspection, and in particular to a method and system for industrial visual inspection based on region of interest fragmentation. Background Technology

[0002] Industrial visual inspection has a wide range of applications in labor-intensive production sectors. Effective detection is a crucial function of such equipment during the visual inspection process. Traditional visual inspection methods mostly employ single region of interest (ROI) detection. While these methods can address defect detection to some extent, they often result in high false negative and over-detection rates, leading to wasted production resources. Modern visual inspection processes require equipment capable of controlling false negative and over-detection rates to a relatively low level, preventing a large number of defective products from mixing with good products or rendering good products unusable.

[0003] Single region of interest (ROI) detection methods employ various localization approaches. For example, template-matching-based ROI localization often presents several problems: First, the time consumption associated with template matching is significant; second, image redundancy within a single ROI is a persistent issue, making it difficult to optimize the method and resulting in suboptimal detection metrics; third, redundant data significantly increases the processing time of the detection algorithm. These issues are unacceptable to users of the equipment.

[0004] In conclusion, finding an accurate and rapid industrial visual inspection method is an urgent problem to be solved. Summary of the Invention

[0005] The first technical problem to be solved by the present invention is to provide an industrial visual inspection method for region-of-interest fragmentation that can quickly and accurately detect defects in a fragmented manner, in contrast to the above-mentioned prior art.

[0006] The second technical problem to be solved by the present invention is to provide an industrial vision inspection system for regions of interest fragmentation that can quickly and accurately detect defects efficiently, in contrast to the above-mentioned prior art.

[0007] The technical solution adopted by the present invention to solve the first technical problem mentioned above is: an industrial visual inspection method for fragmented regions of interest, characterized by comprising the following steps:

[0008] Step 1, offline build steps, specifically including:

[0009] Step 1.1: Obtain an offline scene grayscale image containing a standard workpiece image;

[0010] Step 1.2: Determine the primary feature targets for coarse positioning from the offline scene grayscale image containing the standard workpiece image, and establish an initial coordinate system. The primary feature target object includes one or any combination of the following features: circle, waist-shaped hole, and hollowed-out long side;

[0011] Step 1.3: Based on the established initial coordinate system In the offline scene grayscale image containing the standard workpiece image, secondary feature targets in the standard workpiece image are identified, and then an offline secondary positioning bounding box is established. 1. Using offline secondary positioning frames 1. Establish the workpiece coordinate system ;

[0012] Step 1.4: Calculate the initial coordinate system With workpiece coordinate system Transformation matrix ;

[0013] Step 1.6: Based on the workpiece coordinate system The set of offline regions of interest for the standard workpiece in the offline scene grayscale image containing images of the standard workpiece. 1. Determine the coordinate system, then calculate the workpiece coordinate system. Collection of offline interest areas The set of transformation matrices of 1 ;

[0014] Step 2, online build steps, specifically including:

[0015] Step 2.1: Obtain a real-time scene grayscale image containing the image of the workpiece to be tested;

[0016] Step 2.2: Determine the primary feature targets for coarse positioning from the real-time scene grayscale image containing the image of the workpiece to be measured, and establish a primary coordinate system. ;

[0017] Step 2.3: Based on the established first-level coordinate system and transformation matrix Offline secondary positioning frame 1. Obtain the real-time secondary positioning frame of the workpiece under test through inverse mapping. ;

[0018] Step 2.4: Utilize the real-time secondary positioning frame of the workpiece to be measured. Establish a second-level coordinate system ;

[0019] Step 2.5: Using a secondary coordinate system and the set of transformation matrices Collect offline interest areas 1. Inverse mapping yields the real-time region of interest set of the workpiece under test. ;

[0020] Step 3: Based on the real-time region of interest set It performs a cropping operation on a real-time scene grayscale image containing an image of the workpiece to be tested.

[0021] As an improvement, in step 3, when performing a cropping operation on the real-time scene grayscale image containing the image of the workpiece to be tested, the following cases are handled respectively:

[0022] If the grayscale is uniform within the real-time region of interest and the grayscale span is less than the preset threshold, select the global segmentation operation process;

[0023] If the grayscale span within the real-time region of interest is greater than or equal to a preset threshold, select image flat field correction or local segmentation method;

[0024] The grayscale values ​​within the real-time region of interest exhibit directionality, allowing for the selection of image convolution or image filtering segmentation methods.

[0025] Further improvements include using a transition point algorithm based on Sobel differentiation to establish the workpiece coordinate system in steps 1.3 and 2.4. and second-level coordinate system .

[0026] Further improvements were made to step 1.3, including the workpiece coordinate system. The specific methods for establishing this include:

[0027] A. Let A be the offline scene grayscale image containing the standard workpiece image. Set the preset convolution template to X and Y. The horizontal edge image is represented as: Ax = X·A. The vertical edge image can be represented as: Ay = Y·A.

[0028] The expressions for the preset convolution templates X and Y are as follows:

[0029] , ;

[0030] B. Transition point calculation: Offline secondary positioning frame The midline of image A is determined by the coordinates of the four corner points of the positioning frame. The midline is denoted as L. Image A is located in the offline secondary positioning frame. The sub-image within 1 is represented as ASUB. A Sobel convolution is performed on ASUB to obtain a convolutional image. The edge line ASUBX of the convolutional image is then obtained, and the offline secondary localization box is generated. The intersection point of the centerline L and the edge line ASUBX of line 1 is calculated to obtain the transition point.

[0031] C. Form a set of transition points, and fit the horizontal point set using the least squares method to construct the workpiece coordinate system. The X-axis and the set of points along the vertical direction are fitted using the least squares method to form the workpiece coordinate system. The Y-axis is used to form the workpiece coordinate system. ;

[0032] Second-level coordinate system The method for establishing the workpiece coordinate system The method for establishing it is the same. In the above steps, replace "offline scene grayscale image containing standard workpiece image" with "real-time scene grayscale image containing workpiece image to be tested", and replace "offline secondary positioning box" with "offline secondary positioning box". Replace "1" with "Real-time secondary positioning box" , to the workpiece coordinate system Replace "secondary coordinate system" with "secondary coordinate system" ".

[0033] Further improvements are made by performing electromechanical image stabilization before steps 1 and 2, and then acquiring an offline scene grayscale image containing a standard workpiece image and a real-time scene grayscale image containing an image of the workpiece to be tested.

[0034] In a further improvement, step 1.4 involves calculating the initial coordinate system using a Cartesian coordinate system transformation. With workpiece coordinate system Transformation matrix Specifically, it includes the following steps:

[0035] Initial coordinate system The homogeneous coordinates are expressed as (x, y, 1), and the workpiece coordinate system is... If the homogeneous coordinates are expressed as (x', y', 1), then a transformation relationship exists:

[0036] ;

[0037] In the formula, θ is the workpiece coordinate system. Relative to the initial coordinate system The rotation angles, Δx and Δy, are the workpiece coordinate system. Relative to the initial coordinate system Translation in the x-direction and translation in the y-direction;

[0038] The initial coordinate system is calculated using the above transformation relationship. With workpiece coordinate system Transformation matrix .

[0039] Further improvements involve performing a rectangular construction operation on the location of any primary feature object, setting a transition point based on grayscale increase or decrease, and defining this as the initial coordinate system. Origin: Based on the tilt angle information of the constructed rectangle relative to the horizontal direction of the real-time scene grayscale image in the offline scene grayscale image and the origin information, an initial coordinate system is constructed. The normal and tangential axes are used to establish the initial coordinate system. .

[0040] The technical solution adopted by the present invention to solve the second technical problem mentioned above is: an industrial visual inspection system for fragmented regions of interest, characterized in that: an inspection device for placing standard workpieces or workpieces to be tested, an industrial camera installed on the inspection device, and a computer-readable storage medium connected to the industrial camera, the computer-readable storage medium storing a computer program that can be read and executed by a processor, wherein the computer program, when executed by the processor, implements the above-mentioned industrial visual inspection method for fragmented regions of interest.

[0041] Compared with the prior art, the advantages of the present invention are: the initial coordinate system is calculated and saved through an offline construction process. With workpiece coordinate system Transformation matrix Construct the workpiece coordinate system Collection of offline interest areas The set of transformation matrices of 1 In the specific online construction steps, the real-time set of regions of interest can be calculated quickly. Furthermore, each real-time scene grayscale image is independently located after acquisition, making it less affected by other factors and achieving high positioning accuracy. This overcomes the problems of redundant information and difficulty in designing detection algorithms caused by a single region of interest. Attached Figure Description

[0042] Figure 1 This is a flowchart of the industrial visual inspection method for fragmented regions of interest in this invention.

[0043] Figure 2 This is a diagram showing the selection of primary feature targets and the calculation of transition points in an offline scene grayscale image in this embodiment of the invention;

[0044] Figure 3 This is a diagram showing the establishment of the initial coordinate system during the offline construction process in this embodiment of the invention;

[0045] Figure 4 This is a diagram showing the frame construction and transition point calculation of the normal and tangential axes of the initial coordinate system during the offline construction process in this embodiment of the invention.

[0046] Figure 5 This is a diagram showing the establishment of the workpiece coordinate system during the offline construction process in this embodiment of the invention;

[0047] Figure 6 This refers to the construction of the offline region of interest set during the offline construction process in this embodiment of the invention;

[0048] Figure 7This is a schematic diagram illustrating the selection of primary feature targets and the calculation of transition points during the online construction process of this invention.

[0049] Figure 8 This is a schematic diagram illustrating the calculation of a second-level coordinate system using an inverse transformation of the first-level coordinate system during the online construction process of an embodiment of the present invention.

[0050] Figure 9 This is a schematic diagram illustrating the real-time region of interest set calculated using the inverse transformation of the second-level coordinate system during the online construction process of this invention.

[0051] Figure 10 This is a schematic diagram of a real-time scene grayscale image containing an image of the workpiece to be tested after cropping in an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the following description, in conjunction with the accompanying drawings, illustrates an industrial visual inspection method and system for region-of-interest fragmentation, using an example of industrial visual inspection of a lead frame. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0053] The industrial visual detection method for region of interest fragmentation according to embodiments of the present invention, such as... Figure 1 As shown, it includes the following steps:

[0054] S100, based on electromechanical image stabilization, acquire an offline scene grayscale image containing a standard lead frame image; in this step, an industrial camera installed on the inspection equipment can be used to acquire a single-channel grayscale image of the scene containing a standard lead frame, so that the next step can continue to extract and analyze the pattern in the image.

[0055] S200, determine the primary feature targets for coarse localization from the offline scene grayscale image containing the standard lead frame image, and establish an initial coordinate system. Based on the established initial coordinate system In the offline scene grayscale image containing the standard lead frame image, secondary feature targets in the standard lead frame image are identified, and then an offline secondary positioning bounding box is established. 1. Using offline secondary positioning frames 1. Establish the workpiece coordinate system Calculate the initial coordinate system With workpiece coordinate system Transformation matrix Based on the workpiece coordinate system The set of offline regions of interest for the standard workpiece in the offline scene grayscale image containing images of the standard workpiece. 1. Determine the coordinate system, then calculate the workpiece coordinate system. Collection of offline interest areas The set of transformation matrices of 1 ;

[0056] S300: Based on electromechanical image stabilization, acquire a real-time scene grayscale image containing the image of the lead frame to be tested; determine the primary feature target object for coarse localization from the real-time scene grayscale image containing the image of the lead frame to be tested, and establish a primary coordinate system. Based on the established first-level coordinate system and transformation matrix Offline secondary positioning frame 1. Obtain the real-time secondary positioning frame of the workpiece under test through inverse mapping. ; Utilizing the real-time two-level positioning frame of the workpiece under test Establish a second-level coordinate system Using a second-level coordinate system and the set of transformation matrices Collect offline interest areas 1. Inverse mapping yields the real-time region of interest set of the workpiece under test. ;

[0057] S400, based on real-time region of interest sets The real-time scene grayscale image containing the image of the lead frame to be tested is cropped. When cropping the real-time scene grayscale image containing the image of the lead frame to be tested, the following procedures apply:

[0058] If the grayscale is uniform within the real-time region of interest and the grayscale span is less than the preset threshold, select the global segmentation operation process;

[0059] If the span within the real-time region of interest is greater than or equal to a preset threshold, select image flat field correction or local segmentation method;

[0060] The grayscale composition within the real-time region of interest is directional, allowing for the selection of image convolution or image filtering segmentation methods.

[0061] Specifically, step S2 includes the following steps:

[0062] S2-1, determine the primary feature targets in the offline scene grayscale image containing the standard lead frame image, construct a rectangular box for the primary feature targets, and crop the sub-images within the rectangular box; the primary feature targets can be one of the following features or any combination thereof: circle, waist-shaped hole, and hollowed-out long side; the secondary feature targets can be one of the following features or any combination thereof: circle, waist-shaped hole, and hollowed-out long side;

[0063] S2-2, For the cropped sub-image rectangle containing the primary feature object, the transition point is obtained using a transition point calculation method based on gray-level increase or decrease, and this transition point is defined as the initial coordinate system. The origin is used to construct an initial coordinate system based on the tilt angle of the cropped sub-image rectangle relative to the horizontal direction in the offline scene grayscale image and the origin information. The normal and tangential axes are used to establish the initial coordinate system. ;

[0064] S2-3, in the offline scene grayscale image containing the standard workpiece image, determine the secondary feature targets in the standard workpiece image, and then establish an offline secondary positioning bounding box. 1. For offline secondary positioning frames Crop the sub-image within 1;

[0065] S2-4: For three cropped sub-images containing secondary feature objects at different positions, obtain the offline secondary localization box. 1. Using the transition point calculation method based on grayscale increase or decrease, obtain the transition points of three offline secondary positioning frames, and construct the workpiece coordinate system using these three transition points. Specifically, the offline scene grayscale image containing the standard workpiece image is represented as A, the preset convolution template is set as X, Y, the horizontal edge image is represented as: Ax = X·A, and the vertical edge image can be represented as: Ay = Y·A;

[0066] The expressions for the preset convolution templates X and Y are as follows:

[0067] , ;

[0068] Offline secondary positioning frame The midline of image A is determined by the coordinates of the four corner points of the positioning frame. The midline is denoted as L. Image A is located in the offline secondary positioning frame. The sub-image within 1 is represented as ASUB. A Sobel convolution is performed on ASUB to obtain a convolutional image. The edge line ASUBX of the convolutional image is then obtained, and the offline secondary localization box is generated. The intersection point of the centerline L and the edge line ASUBX of line 1 is calculated to obtain the transition point.

[0069] C. Form a set of transition points, and fit the horizontal point set using the least squares method to construct the workpiece coordinate system. The X-axis and the set of points along the vertical direction are fitted using the least squares method to form the workpiece coordinate system. The Y-axis is used to form the workpiece coordinate system. .

[0070] S2-5, Calculate the initial coordinate system With workpiece coordinate system Transformation matrix ;

[0071] S2-6, based on class criteria and workpiece coordinate system The set of offline regions of interest for the standard workpiece in the offline scene grayscale image containing images of the standard workpiece. 1. Determine the coordinate system, then calculate the workpiece coordinate system. Collection of offline interest areas The set of transformation matrices of 1 .

[0072] like Figure 2 The diagram shows a schematic of the primary feature object selected for the initial coordinate system. The left side is a grayscale image of a workpiece; the right side shows the relationship between the primary feature object and the initial bounding box. Clearly, a bounding box is selected here at the boundary between the workpiece and the background, and the arrow indicates that the transition point calculation is performed using a positive method.

[0073] like Figure 3 As shown, an initial coordinate system is constructed based on this point, and the black coordinate system diagram is the initial coordinate system.

[0074] like Figure 4 As shown, based on the initial coordinate system, three boxes are constructed, located at the junction of the foreground and background of their respective secondary feature objects. The arrow symbol indicates that the transition point calculation is performed in a positive manner, and the transformation relationship M1 between the initial coordinate system and the three boxes is obtained.

[0075] like Figure 5 As shown, the workpiece coordinate system is constructed based on these three points, and the gray coordinate system diagram is the workpiece coordinate system.

[0076] like Figure 6 As shown in the figure, the right box shows the set of regions of interest for a certain type of detection requirement, and the transformation relationship M2 between the workpiece coordinate system and the regions of interest is calculated.

[0077] like Figure 7 As shown, the initial coordinate system of the real-time image is established using the initial feature bounding box.

[0078] like Figure 8 As shown, the workpiece coordinate system in the real-time image is established using the transformation relationship M1 between the initial coordinate system and the secondary feature target object frame.

[0079] like Figure 9 As shown, the inverse transformation operation of the region of interest in the real-time image utilizes the transformation relationship M2 between the workpiece coordinate system and the region of interest.

[0080] This system can fragment and precisely locate regions of interest in workpiece inspection, and perform class-based inspection. Furthermore, the hardware configuration is simple, requiring only the industrial camera to be mounted in a suitable position on the inspection machine. The host computer can be a common device with computing capabilities, such as a general-purpose computer, or a microcontroller capable of image processing. The industrial camera can be positioned at the top or bottom of the equipment depending on the actual operating scenario.

[0081] Preferably, in other embodiments, an LED light source can also be provided to complement the industrial camera 100. The LED light source is also mounted on the device, generally close to the industrial camera 100, so that the industrial camera can acquire a clearer scene image.

[0082] The industrial camera installed on the equipment can be an area scan camera, a line scan camera, a USB interface camera, a 1394 interface camera, a network communication camera, or a webcam. A suitable industrial camera can be selected based on the host computer's interface and cost requirements, as long as it can acquire scene images.

[0083] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method of interest area fragmentation industrial vision detection, characterized in that Includes the following steps: Step 1, offline build steps, specifically including: Step 1.1: Obtain an offline scene grayscale image containing a standard workpiece image; Step 1.

2. Determine the coarse positioning primary feature target from the offline scene gray image containing standard workpiece image, establish initial coordinate system , the primary feature target contains one or any combination of the following features: circle, waist hole, hollow long side; Step 1.3, establishing the initial coordinate system based on the established initial coordinate system In the offline scene gray image containing the standard workpiece image, the secondary feature target in the standard workpiece image is determined, and then an offline secondary positioning frame is established 1, using the offline secondary positioning frame 1establishing a workpiece coordinate system ; Step 1.4, calculating the initial coordinate system with the workpiece coordinate system conversion matrix ; Step 1.5, determining a workpiece coordinate system based on the workpiece coordinate system a set of offline regions of interest of the standard workpiece in the offline scene grayscale image containing the standard workpiece image 1, and then calculating a workpiece coordinate system and a set of transformation matrices of the set of offline regions of interest 1 ; Step 2, online build steps, specifically including: Step 2.1: Obtain a real-time scene grayscale image containing the image of the workpiece to be tested; Step 2.2: Determine the primary feature targets for coarse positioning from the real-time scene grayscale image containing the image of the workpiece to be measured, and establish a primary coordinate system. ; Step 2.3: Based on the established first-level coordinate system and transformation matrix Offline secondary positioning frame 1. Obtain the real-time secondary positioning frame of the workpiece under test through inverse mapping. ; Step 2.4: Utilize the real-time secondary positioning frame of the workpiece to be measured. Establish a second-level coordinate system ; Step 2.5: Using a secondary coordinate system and the set of transformation matrices Collect offline interest areas 1. Inverse mapping yields the real-time region of interest set of the workpiece under test. ; Step 3: Based on the real-time region of interest set It performs a cropping operation on a real-time scene grayscale image containing an image of the workpiece to be tested.

2. The industrial visual inspection method for fragmented regions of interest according to claim 1, characterized in that: In step 3, when performing a cropping operation on the real-time scene grayscale image containing the image of the workpiece to be tested, the following cases are handled respectively: If the grayscale is uniform within the real-time region of interest and the grayscale span is less than the preset threshold, select the global segmentation operation process; If the grayscale span within the real-time region of interest is greater than or equal to a preset threshold, select image flat field correction or local segmentation method; The grayscale values ​​within the real-time region of interest exhibit directionality, allowing for the selection of image convolution or image filtering segmentation methods.

3. The industrial visual inspection method for fragmented regions of interest according to claim 1, characterized in that: In steps 1.3 and 2.4, the workpiece coordinate system is established using a transition point algorithm based on Sobel differentiation. and second-level coordinate system .

4. The industrial visual inspection method for fragmented regions of interest according to claim 3, characterized in that: In step 1.3, the workpiece coordinate system The specific methods for establishing this include: A. Let A be the offline scene grayscale image containing the standard workpiece image. Set the preset convolution template to X and Y. The horizontal edge image is represented as: Ax = X·A. The vertical edge image can be represented as: Ay = Y·A. The expressions for the preset convolution templates X and Y are as follows: ， ； B. Transition point calculation: Offline secondary positioning frame The midline of image A is determined by the coordinates of the four corner points of the positioning frame. The midline is denoted as L. Image A is located in the offline secondary positioning frame. The sub-image within 1 is represented as ASUB. A Sobel convolution is performed on ASUB to obtain a convolutional image. The edge line ASUBX of the convolutional image is then obtained, and the offline secondary localization box is generated. The intersection point of the centerline L and the edge line ASUBX of line 1 is calculated to obtain the transition point. C. Form a set of transition points, and fit the horizontal point set using the least squares method to construct the workpiece coordinate system. The X-axis and the set of points along the vertical direction are fitted using the least squares method to form the workpiece coordinate system. The Y-axis is used to form the workpiece coordinate system. ; Second-level coordinate system The method for establishing the workpiece coordinate system The method for establishing it is the same. In the above steps, replace "offline scene grayscale image containing standard workpiece image" with "real-time scene grayscale image containing workpiece image to be tested", and replace "offline secondary positioning box" with "offline secondary positioning box". Replace "1" with "Real-time secondary positioning box" , to the workpiece coordinate system Replace with "secondary coordinate system" ".

5. The industrial visual inspection method for fragmented regions of interest according to claim 1, characterized in that: Before executing steps 1 and 2, electromechanical image stabilization is performed first, and then an offline scene grayscale image containing a standard workpiece image and a real-time scene grayscale image containing an image of the workpiece to be tested are acquired.

6. The industrial visual inspection method for fragmented regions of interest according to claim 1, characterized in that: In step 1.4, the initial coordinate system is calculated using a Cartesian coordinate system transformation. With workpiece coordinate system Transformation matrix Specifically, it includes the following steps: Initial coordinate system The homogeneous coordinates are expressed as (x, y, 1), and the workpiece coordinate system is... If the homogeneous coordinates are expressed as (x', y', 1), then a transformation relationship exists: ； In the formula, θ is the workpiece coordinate system. Relative to the initial coordinate system The rotation angles, Δx and Δy, are the workpiece coordinate system. Relative to the initial coordinate system Translation in the x-direction and translation in the y-direction; The initial coordinate system is calculated using the above transformation relationship. With workpiece coordinate system Transformation matrix .

7. The industrial visual inspection method for fragmented regions of interest according to claim 1, characterized in that: Perform a rectangular construction operation on the location of any primary feature object, set a transition point based on grayscale increase or decrease, and define this as the initial coordinate system. Origin: Based on the tilt angle information of the constructed rectangle relative to the horizontal direction of the real-time scene grayscale image in the offline scene grayscale image and the origin information, an initial coordinate system is constructed. The normal and tangential axes are used to establish the initial coordinate system. .

8. An industrial vision inspection system for fragmented regions of interest, characterized in that: An inspection device for placing standard workpieces or workpieces to be tested, an industrial camera mounted on the inspection device, and a computer-readable storage medium connected to the industrial camera, the computer-readable storage medium storing a computer program that can be read and executed by a processor, wherein the computer program, when executed by the processor, implements the region-of-interest fragmentation industrial visual inspection method as described in claim 1.

Images