A high-speed visual inspection method for surface defects of microelectronic components

Abstract

The application discloses a high-speed visual detection method for surface defects of micro electronic components, and relates to the technical field of microelectronics, which comprises the following steps: setting high-speed cameras at the upper, lower, left and right positions of a transparent conveying belt; forming at least one sample splicing image, and analyzing to obtain sample pixel value distribution of the sample splicing image; forming a preset time; obtaining a first component area, and analyzing to obtain first pixel value distribution of the first component area; obtaining a second component area, and analyzing to obtain second pixel value distribution of the second component area; screening a first target image from the first image, and screening a second target image from the second image; and judging whether the electronic component has defects or not. Through the steps of setting the positions of the high-speed cameras, forming the sample pixel value distribution, screening the first target image and screening the second target image, the defects can be accurately recognized, and the calculation amount for detection can be effectively limited.

Description

Technical Field

[0001] This invention relates to the field of microelectronics technology, and more specifically to a high-speed visual inspection method for surface defects in microelectronic components. Background Technology

[0002] Microelectronic components are the fundamental units that make up modern electronic devices. Through micrometer / nanometer-level manufacturing techniques, circuits are miniaturized to achieve electrical functions in extremely small sizes. Various unforeseen circumstances can arise during the production of microelectronic components, leading to problems such as pin abnormalities, surface scratches, or cracks. However, during inspection, the orientation of microelectronic components cannot be consistent, and the six surfaces to be inspected—top, bottom, left, right, front, and back—are uncertain due to their unknown orientation, resulting in an unpredictable size for each surface in the image. Furthermore, the presence of defects further complicates accurate comparison and inspection, making accurate defect identification difficult with current technologies. Summary of the Invention

[0003] To address the aforementioned technical problems, a high-speed visual inspection method for surface defects in microelectronic components is provided. This technical solution solves the problems mentioned in the background section.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A high-speed visual inspection method for surface defects in microelectronic components includes: A transparent conveyor belt is set up to transport electronic components. The electronic components are pushed by a robot to the center line of the transparent conveyor belt for transport. The space where the transparent conveyor belt is located is a pure white space. High-speed cameras are installed at the top, bottom, left, and right positions of the transparent conveyor belt to acquire qualified historical electronic components in advance as sample components. Pre-acquire sample images of the sample element from at least one perspective, wherein the sample images are images acquired from the top, bottom, left, right, front, and rear perspectives of the sample element; Based on the sample images, at least one sample stitched image is formed, and the distribution of sample pixel values ​​in the sample stitched image is analyzed. A preset time is set, and at each preset time interval, a high-speed camera at the top and bottom positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one first image, and a high-speed camera at the left and right positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one second image. The electronic components in the first image are identified to obtain the first component region, and the distribution of the first pixel value in the first component region is analyzed. The electronic components in the second image are identified to obtain the second component region, and the distribution of the second pixel values ​​in the second component region is analyzed. Based on the first pixel value distribution and the sample pixel value distribution, a first target image is obtained by filtering from the first image; based on the second pixel value distribution and the sample pixel value distribution, a second target image is obtained by filtering from the second image. Based on the first target image and the second target image, determine whether there are defects in the electronic components.

[0005] Preferably, the step of installing high-speed cameras at the top, bottom, left, and right positions of the transparent conveyor belt includes the following steps: High-speed cameras are installed at both the top and bottom of the transparent conveyor belt, and the high-speed cameras at the top and bottom are located in the vertical plane of the center line of the transparent conveyor belt; The high-speed camera at the upper position is not within the field of view of the high-speed camera at the lower position, and vice versa. High-speed cameras are installed on both the left and right sides of the transparent conveyor belt, and the high-speed cameras on the left and right sides are located in the plane of the transparent conveyor belt; The high-speed camera on the left is not within the field of view of the high-speed camera on the right, and the high-speed camera on the right is not within the field of view of the high-speed camera on the left. The left and right positions refer to the two sides of the transparent conveyor belt's forward direction.

[0006] Preferably, forming at least one sample stitched image based on the sample images includes the following steps: The sample image is used as the sample stitched image, and the sample images obtained from adjacent viewpoints are stitched together to obtain the sample stitched image.

[0007] Preferably, the analysis of the sample pixel value distribution of the stitched image includes the following steps: Each of the different pixel values ​​contained in the stitched image is taken as a sample pixel value. The number of pixels in the stitched image whose pixel value is equal to the sample pixel value is counted and taken as the sample count. Divide the number of samples by the total number of pixels in the stitched image to obtain the proportion of sample pixel values. Use the proportion of sample pixel values ​​and the sample pixel values ​​as the distribution of sample pixel values ​​in the stitched image.

[0008] Preferably, the formation of the preset time includes the following steps: The allowable error ratio of visual inspection is obtained in advance, and at least one reference time is randomly generated, with the reference time not exceeding 1 second; Using a reference time interval, a high-speed camera at an upper position is used to acquire images of electronic components transported by a transparent conveyor belt, resulting in front and rear images; The pixel value of the white pixel is taken as the white pixel value; Pixels in the foreground image whose pixel value is not equal to the white pixel value are taken as foreground pixels. The foreground pixels are aggregated to obtain the foreground region. The minimum distance from the point in the foreground region to the right edge of the foreground image is taken as the foreground distance. Pixels in the rear image whose pixel value is not equal to the white pixel value are taken as rear pixels. The rear pixels are aggregated to obtain the rear region. The minimum distance from the point in the rear region to the right edge of the rear image is taken as the rear distance. Divide the absolute value of the difference between the front distance and the back distance by the distance between the left and right edges of the front image to obtain the reference error ratio; If the reference error ratio does not exceed the allowable error ratio, the reference time with the reference error ratio will be used as the target time, and the remaining reference times will be used as non-target times. The minimum value of the non-target time is used as the preset value, and the maximum value of at least one target time that does not exceed the preset value is used as the preset time.

[0009] Preferably, obtaining the first element region and analyzing the distribution of the first pixel values ​​in the first element region includes the following steps: The pixels in the first image whose pixel value is not equal to the white pixel value are taken as the first pixel, and the first pixels are aggregated to form the first element region; Each of the different pixel values ​​contained in the first element region is taken as the first pixel value. The number of pixels in the first element region whose pixel value is equal to the first pixel value is counted and taken as the first number. Divide the first number by the total number of pixels in the first element region to obtain the proportion of the first pixel value. Use the proportion of the first pixel value and the first pixel value as the distribution of the first pixel value in the first element region.

[0010] Preferably, obtaining the second element region and analyzing the distribution of the second pixel values ​​in the second element region includes the following steps: In the second image, pixels whose pixel values ​​are not equal to white pixel values ​​are taken as second pixels, and these second pixels are aggregated to form a second element region. Each of the different pixel values ​​contained in the second element region is taken as the second pixel value. The number of pixels in the second element region whose pixel value is equal to the second pixel value is counted and taken as the second number. Divide the second number by the total number of pixels in the second element region to obtain the proportion of the second pixel value. Use the proportion of the second pixel value and the second pixel value as the distribution of the second pixel value in the second element region.

[0011] Preferably, obtaining the first target image from the first image includes the following steps: The first image with the smallest number of first pixel values ​​is selected as the first target image.

[0012] Preferably, obtaining the second target image from the second image includes the following steps: Establish a correspondence between the second pixel value in the second image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the second pixel value in the second image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the second image relative to the sample stitched image. The minimum error value of the second image relative to at least one sample stitched image is taken as the actual error ratio of the second image, and the second image with the smallest actual error ratio is taken as the second target image.

[0013] Preferably, determining whether an electronic component has a defect includes the following steps: Establish a correspondence between the first pixel value in the first target image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the first pixel value in the first target image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the first target image relative to the sample stitched image. The minimum error value of the first target image relative to at least one sample stitched image is taken as the actual error ratio of the first target image; If the actual error ratio of the first target image and the actual error ratio of the second target image both do not exceed the allowable error ratio, then the electronic component is not defective; otherwise, the electronic component is defective.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: By setting the position of a high-speed camera, generating a sample pixel value distribution, and filtering to obtain a first target image and a second target image, images of electronic components from different perspectives can be acquired through the setting of high-speed cameras with different viewpoints. Based on the analysis of pixel value proportions, the desired image is selected from the acquired images, and the detection result is obtained based on the pixel value proportions of the filtered images. This avoids situations where the pose and size of electronic components in the detected images are unknown, thus enabling more accurate defect identification while effectively limiting the computational load used for detection. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the high-speed visual inspection method for surface defects in microelectronic components according to the present invention. Figure 2This is a schematic diagram illustrating the process of installing high-speed cameras at the top, bottom, left, and right positions of the transparent conveyor belt according to the present invention. Figure 3 This is a schematic diagram illustrating the process of obtaining the sample pixel value distribution of the stitched image according to the present invention. Figure 4 This is a schematic diagram of the process for forming a preset time according to the present invention; Figure 5 This is a schematic diagram illustrating the process of analyzing the distribution of the first pixel value in the first element region to obtain the first element region according to the present invention. Figure 6 This is a schematic diagram illustrating the process of analyzing the distribution of second pixel values ​​in the second element region to obtain the second element region according to the present invention. Figure 7 This is a schematic diagram of the process of obtaining a second target image from a second image according to the present invention; Figure 8 This is a schematic diagram of the process for determining whether an electronic component has a defect according to the present invention. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] Reference Figure 1 As shown, a high-speed visual inspection method for surface defects in microelectronic components includes: A transparent conveyor belt is set up to transport electronic components. The electronic components are pushed by a robot to the center line of the transparent conveyor belt for transport. The space where the transparent conveyor belt is located is a pure white space. High-speed cameras are installed at the top, bottom, left, and right positions of the transparent conveyor belt to acquire qualified historical electronic components in advance as sample components. Pre-acquire sample images of the sample element from at least one perspective, wherein the sample images are images acquired from the top, bottom, left, right, front, and rear perspectives of the sample element; Based on the sample images, at least one sample stitched image is formed, and the distribution of sample pixel values ​​in the sample stitched image is analyzed. A preset time is set, and at each preset time interval, a high-speed camera at the top and bottom positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one first image, and a high-speed camera at the left and right positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one second image. The electronic components in the first image are identified to obtain the first component region, and the distribution of the first pixel value in the first component region is analyzed. The electronic components in the second image are identified to obtain the second component region, and the distribution of the second pixel values ​​in the second component region is analyzed. Based on the first pixel value distribution and the sample pixel value distribution, a first target image is obtained by filtering from the first image; based on the second pixel value distribution and the sample pixel value distribution, a second target image is obtained by filtering from the second image. Based on the first target image and the second target image, determine whether there are defects in the electronic components.

[0018] In this scheme, the main focus is on detecting microelectronic components that are approximately cuboid in shape. During detection, six perspectives need to be examined: front and back, left and right, and top and bottom. This method can also be used to detect other microelectronic components, but the accuracy will be slightly lower because micro-resistors are cylindrical, which is quite different from cuboids. Microelectronic components that are approximately cuboid in shape include microchips and other electronic components that form a cuboid. The six faces of the cuboid are numbered as follows: the top is 1, the bottom face (the face in contact with the conveyor belt) is 6, and the sides are numbered 2, 3, 4, and 5. The following descriptions of the technology will use this cuboid as the reference point for the locations where electronic components appear. Existing technologies for inspecting cuboids primarily rely on image capture, comparing the acquired images with defect-free sample images to determine the presence of defects. However, this method presents significant challenges because the cuboid's orientation is uncertain. For instance, faces 2 and 3 may face the camera, but the angles between faces 2 and 3 and the shooting angle are unpredictable, resulting in inconsistent dimensions for faces 2 and 3 in the image. This necessitates separate processing of faces 2 and 3, requiring individual identification and differentiation. However, due to the unknown dimensions of faces 2 and 3 and the potential presence of defects, identification becomes extremely difficult. Therefore, existing comparison methods are computationally intensive, demanding substantial computational resources, and their accuracy needs improvement. Here, the shooting angle is the line connecting the high-speed camera and the cuboid. In order to overcome the problems existing in the current technology, a series of steps are set up to address them.

[0019] Reference Figure 2 As shown, installing high-speed cameras at the top, bottom, left, and right positions of the transparent conveyor belt includes the following steps: High-speed cameras are installed at both the top and bottom of the transparent conveyor belt, and the high-speed cameras at the top and bottom are located in the vertical plane of the center line of the transparent conveyor belt; The high-speed camera at the upper position is not within the field of view of the high-speed camera at the lower position, and vice versa. High-speed cameras are installed on both the left and right sides of the transparent conveyor belt, and the high-speed cameras on the left and right sides are located in the plane of the transparent conveyor belt; The high-speed camera on the left is not within the field of view of the high-speed camera on the right, and the high-speed camera on the right is not within the field of view of the high-speed camera on the left. The left and right positions refer to the two sides of the transparent conveyor belt's forward direction.

[0020] The purpose of this setting is that when no electronic components are present, the high-speed camera will capture white images, making it easy to identify any electronic components that are not white. Since the electronic component is pushed to the center line of the transparent conveyor belt by the robotic arm for transmission, the shooting situation at the top and bottom positions is similar, and the shooting situation at the left and right positions is similar. Therefore, taking the top and left positions as examples, it is easy to know that when the electronic component is not directly below the top position, the top position will shoot three faces. Taking a cuboid as an example, it can be faces numbered 1, 2, and 3, or it can be faces numbered 1 and 2. The former is when all the edges of the cuboid are not parallel to the center line of the transparent conveyor belt, and the latter is when four edges of the cuboid are parallel to the center line of the transparent conveyor belt. When the electronic component is directly below the top position, the face shot is face numbered 1. Since the high-speed camera on the left is on the same plane as the transparent conveyor belt, it only photographs the faces numbered 2, 3, 4 and 5. Depending on the orientation of the cuboid, it could be the faces numbered 2 and 3, the faces numbered 3 and 4, the faces numbered 4 and 5, or the faces numbered 5 and 2. In short, it will only photograph two faces. It is easy to know that the images taken from the top, bottom, left, and right positions contain images of all faces of the cuboid. Therefore, as long as the images obtained are free of defects, the cuboid must be free of defects. The above analysis of the photographed surfaces will be used for defect detection and identification in the future.

[0021] Creating at least one sample stitched image based on the sample images includes the following steps: The sample image is used as the sample stitched image, and the sample images obtained from adjacent viewpoints are stitched together to obtain the sample stitched image.

[0022] Here, the sample stitched images are used for subsequent comparison. Since the images taken from the top and bottom positions will only select to photograph the face numbered 1 or 6 in the subsequent comparison, and the images taken from the left and right positions only include two faces, the obtained images only contain one face of the cuboid or the image of two adjacent faces. Therefore, the images used for comparison only need to include the case of one face and two adjacent faces. Here, the adjacent viewpoints are front view and left view, front view and up view, front view and right view, front view and down view, back view and left view, back view and up view, back view and right view, back view and down view, down view and left view, down view and right view, up view and left view, up view and right view; Although the images taken from the left and right positions are only images of two adjacent sides of the cuboid, since the cuboid's orientation is unknown, its bottom face may correspond to any of the sample images from the six perspectives. Therefore, the adjacent perspectives include the above situations.

[0023] Reference Figure 3 As shown, the analysis of the sample pixel value distribution of the stitched image includes the following steps: Each of the different pixel values ​​contained in the stitched image is taken as a sample pixel value. The number of pixels in the stitched image whose pixel value is equal to the sample pixel value is counted and taken as the sample count. Divide the number of samples by the total number of pixels in the stitched image to obtain the proportion of sample pixel values. Use the proportion of sample pixel values ​​and the sample pixel values ​​as the distribution of sample pixel values ​​in the stitched image.

[0024] In this scheme, defects are mainly detected by the proportion of pixel values. This eliminates the need to process the size of the image. It is easy to see that the proportion of different pixel values ​​in the image of a surface remains unchanged regardless of the shooting angle or shooting distance. This principle will be used for defect identification in the future.

[0025] Reference Figure 4 As shown, the process of setting a preset time includes the following steps: The allowable error ratio of visual inspection is obtained in advance, and at least one reference time is randomly generated, with the reference time not exceeding 1 second; Using a reference time interval, a high-speed camera at an upper position is used to acquire images of electronic components transported by a transparent conveyor belt, resulting in front and rear images; The pixel value of the white pixel is taken as the white pixel value; Pixels in the foreground image whose pixel value is not equal to the white pixel value are taken as foreground pixels. The foreground pixels are aggregated to obtain the foreground region. The minimum distance from the point in the foreground region to the right edge of the foreground image is taken as the foreground distance. Pixels in the rear image whose pixel value is not equal to the white pixel value are taken as rear pixels. The rear pixels are aggregated to obtain the rear region. The minimum distance from the point in the rear region to the right edge of the rear image is taken as the rear distance. Divide the absolute value of the difference between the front distance and the back distance by the distance between the left and right edges of the front image to obtain the reference error ratio; If the reference error ratio does not exceed the allowable error ratio, the reference time with the reference error ratio will be used as the target time, and the remaining reference times will be used as non-target times. The minimum value of the non-target time is used as the preset value, and the maximum value of at least one target time that does not exceed the preset value is used as the preset time.

[0026] The preset time is mainly used to control detection errors. With sufficient accuracy, the number of images acquired is minimized, thereby reducing the workload of image processing. It should be noted that the reference time is judged based on the ratio of reference errors generated by only two images, which has a certain deviation. That is, a larger reference time may be mistakenly identified as the target time. In fact, as long as more sets of images are acquired, the reference time will be judged as a non-target time. There will inevitably be non-target times smaller than the target time in this case. Therefore, in order to filter out such target times, the maximum value of at least one target time not exceeding the preset value is used as the preset time. Thus, the generated preset time is more reasonable.

[0027] Reference Figure 5 As shown, the first element region is obtained, and the analysis of the first pixel value distribution of the first element region includes the following steps: The pixels in the first image whose pixel value is not equal to the white pixel value are taken as the first pixel, and the first pixels are aggregated to form the first element region; Each of the different pixel values ​​contained in the first element region is taken as the first pixel value. The number of pixels in the first element region whose pixel value is equal to the first pixel value is counted and taken as the first number. Divide the first number by the total number of pixels in the first element region to obtain the proportion of the first pixel value. Use the proportion of the first pixel value and the first pixel value as the distribution of the first pixel value in the first element region.

[0028] Reference Figure 6 As shown, the second element region is obtained, and the analysis of the second pixel value distribution of the second element region includes the following steps: In the second image, pixels whose pixel values ​​are not equal to white pixel values ​​are taken as second pixels, and these second pixels are aggregated to form a second element region. Each of the different pixel values ​​contained in the second element region is taken as the second pixel value. The number of pixels in the second element region whose pixel value is equal to the second pixel value is counted and taken as the second number. Divide the second number by the total number of pixels in the second element region to obtain the proportion of the second pixel value. Use the proportion of the second pixel value and the second pixel value as the distribution of the second pixel value in the second element region.

[0029] The process of selecting the first target image from the first image includes the following steps: The first image with the smallest number of first pixel values ​​is selected as the first target image.

[0030] At least one first image is a set of multiple images of the same electronic component acquired at preset intervals. The second image is similar. In order to ensure that the image contains only one electronic component, the electronic components need to be spaced out. This can be done by a robotic arm. At the same time, the distance between the high-speed camera and the transparent conveyor belt needs to be reduced so that the required image can be acquired with the electronic components spaced out. Based on the previous analysis, taking the above position as an example, the lower position is similar. The image captured by the first image must contain the image numbered 1. It may contain only the image numbered 1, or it may contain images of multiple faces. Since the cuboid corresponds to electronic components, the pixel values ​​contained in its different faces are different. Therefore, the number of different pixel values ​​in the image containing more faces is greater. Therefore, the first target image can be obtained by filtering based on this, which contains only the image numbered 1. Here, we cannot simplify the process by using the same filtering method as for the second image. This is because the second image only contains two adjacent faces, such as 2 and 3, while the first image may contain three faces, such as 1, 2, and 3. Although the cuboid's orientation is unknown, it is fixed. Its shooting angle is the line connecting the high-speed camera and the cuboid. Imaging is achieved through projection along the shooting angle. As the cuboid moves, the shooting angle constantly changes. There must be a moment when the angle between the shooting angle and faces 2 and 3 is the same. Therefore, faces 2 and 3 appear to be the same size in the image because they are projected at the same angle. However, in other cases, the angles between faces 2 and 3 and the shooting angle are different, thus they are projected at different angles. In an image, the dimensions presented are inconsistent. It is easy to see that if two faces are projected at the same angle, the proportion of different pixel values ​​in the resulting composite image remains unchanged, that is, it is consistent with the proportion of the original two composite pixel values. If two faces are projected at different angles, the proportion of different pixel values ​​in the resulting composite image remains unchanged, that is, it is inconsistent with the proportion of the original two composite pixel values. Therefore, the second target image can be obtained by screening based on this. The difference in the proportion of pixel values ​​between the composite image formed by two faces projected at the same angle and the sample stitched image is the smallest. This is satisfied when there are defects. This is the principle for subsequently screening the second target image from the second image. However, when the first image contains faces 1, 2, and 3, there is only one such position when the angle between 2 and 3 and the shooting angle is the same. At this position, the shooting angle may not be the same as that of 1. Therefore, the sizes of 1, 2, and 3 in the image are inconsistent, resulting in a different pixel value distribution than the pixel value distribution of the combined image of the frontal view of 1, 2, and 3. Therefore, other methods are needed to filter the first target image.

[0031] Reference Figure 7 As shown, obtaining the second target image from the second image includes the following steps: Establish a correspondence between the second pixel value in the second image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the second pixel value in the second image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the second image relative to the sample stitched image. The minimum error value of the second image relative to at least one sample stitched image is taken as the actual error ratio of the second image, and the second image with the smallest actual error ratio is taken as the second target image.

[0032] Here, the error value of the second image relative to the sample stitched image is generated by the stitching of a single second image with a single sample image.

[0033] Reference Figure 8 As shown, determining whether an electronic component has a defect includes the following steps: Establish a correspondence between the first pixel value in the first target image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the first pixel value in the first target image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the first target image relative to the sample stitched image. The minimum error value of the first target image relative to at least one sample stitched image is taken as the actual error ratio of the first target image; If the actual error ratio of the first target image and the actual error ratio of the second target image both do not exceed the allowable error ratio, then the electronic component is not defective; otherwise, the electronic component is defective.

[0034] The first target image is generated from the top or bottom position, therefore it contains face 1 or 6. The second target image is generated from the left or right position. If the second target image generated from the left position contains face 2 and face 3, then the second target image generated from the right position must contain face 4 and face 5. Therefore, all the first and second target images contain all the faces of the cuboid. Thus, when their errors are all within the allowable error ratio, it means that there are no defects.

[0035] Furthermore, this solution also proposes a storage medium on which a computer-readable program is stored, which, when invoked, executes the aforementioned high-speed visual inspection method for surface defects of microelectronic components.

[0036] It is understandable that the storage medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid-state drive (SSD).

[0037] In summary, the advantages of this invention are as follows: by setting the position of the high-speed camera, forming a sample pixel value distribution, and filtering to obtain a first target image and a second target image, images of electronic components from different perspectives can be acquired through the setting of high-speed cameras with different viewpoints. Based on the analysis of pixel value proportions, the desired image can be selected from the acquired images, and the detection result can be obtained based on the pixel value proportions of the filtered images. This avoids situations where the pose and size of the electronic components in the detected images are unknown, thus enabling more accurate defect identification while effectively limiting the computational load used for detection.

[0038] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A high-speed visual inspection method for surface defects in microelectronic components, characterized in that, include: A transparent conveyor belt is set up to transport electronic components. The electronic components are pushed by a robot to the center line of the transparent conveyor belt for transport. The space where the transparent conveyor belt is located is a pure white space. High-speed cameras are installed at the top, bottom, left, and right positions of the transparent conveyor belt to acquire qualified historical electronic components in advance as sample components. Pre-acquire sample images of the sample element from at least one perspective, wherein the sample images are images acquired from the top, bottom, left, right, front, and rear perspectives of the sample element; Based on the sample images, at least one sample stitched image is formed, and the distribution of sample pixel values ​​in the sample stitched image is analyzed. A preset time is set, and at each preset time interval, a high-speed camera at the top and bottom positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one first image, and a high-speed camera at the left and right positions takes pictures of the electronic components on the transparent conveyor belt to obtain at least one second image. The electronic components in the first image are identified to obtain the first component region, and the distribution of the first pixel value in the first component region is analyzed. The electronic components in the second image are identified to obtain the second component region, and the distribution of the second pixel values ​​in the second component region is analyzed. Based on the first pixel value distribution and the sample pixel value distribution, a first target image is obtained by filtering from the first image; based on the second pixel value distribution and the sample pixel value distribution, a second target image is obtained by filtering from the second image. Based on the first target image and the second target image, determine whether there are defects in the electronic components.

2. The high-speed visual inspection method for surface defects of microelectronic components according to claim 1, characterized in that, The method of installing high-speed cameras at the top, bottom, left, and right positions of the transparent conveyor belt includes the following steps: High-speed cameras are installed at both the top and bottom of the transparent conveyor belt, and the high-speed cameras at the top and bottom are located in the vertical plane of the center line of the transparent conveyor belt; The high-speed camera at the upper position is not within the field of view of the high-speed camera at the lower position, and vice versa. High-speed cameras are installed on both the left and right sides of the transparent conveyor belt, and the high-speed cameras on the left and right sides are located in the plane of the transparent conveyor belt; The high-speed camera on the left is not within the field of view of the high-speed camera on the right, and the high-speed camera on the right is not within the field of view of the high-speed camera on the left. The left and right positions refer to the two sides of the transparent conveyor belt's forward direction.

3. The high-speed visual inspection method for surface defects of microelectronic components according to claim 2, characterized in that, The process of forming at least one sample stitched image based on the sample image includes the following steps: The sample image is used as the sample stitched image, and the sample images obtained from adjacent viewpoints are stitched together to obtain the sample stitched image.

4. The high-speed visual inspection method for surface defects of microelectronic components according to claim 3, characterized in that, The analysis to obtain the sample pixel value distribution of the stitched image includes the following steps: Each of the different pixel values ​​contained in the stitched image is taken as a sample pixel value. The number of pixels in the stitched image whose pixel value is equal to the sample pixel value is counted and taken as the sample count. Divide the number of samples by the total number of pixels in the stitched image to obtain the proportion of sample pixel values. Use the proportion of sample pixel values ​​and the sample pixel values ​​as the distribution of sample pixel values ​​in the stitched image.

5. A high-speed visual inspection method for surface defects of microelectronic components according to claim 4, characterized in that, The process of setting the preset time includes the following steps: The allowable error ratio of visual inspection is obtained in advance, and at least one reference time is randomly generated, with the reference time not exceeding 1 second; Using a reference time interval, a high-speed camera at an upper position is used to acquire images of electronic components transported by a transparent conveyor belt, resulting in front and rear images; The pixel value of the white pixel is taken as the white pixel value; Pixels in the foreground image whose pixel value is not equal to the white pixel value are taken as foreground pixels. The foreground pixels are aggregated to obtain the foreground region. The minimum distance from the point in the foreground region to the right edge of the foreground image is taken as the foreground distance. Pixels in the rear image whose pixel value is not equal to the white pixel value are taken as rear pixels. The rear pixels are aggregated to obtain the rear region. The minimum distance from the point in the rear region to the right edge of the rear image is taken as the rear distance. Divide the absolute value of the difference between the front distance and the back distance by the distance between the left and right edges of the front image to obtain the reference error ratio; If the reference error ratio does not exceed the allowable error ratio, the reference time with the reference error ratio will be used as the target time, and the remaining reference times will be used as non-target times. The minimum value of the non-target time is used as the preset value, and the maximum value of at least one target time that does not exceed the preset value is used as the preset time.

6. A high-speed visual inspection method for surface defects of microelectronic components according to claim 5, characterized in that, The process of obtaining the first element region and analyzing the distribution of the first pixel values ​​in the first element region includes the following steps: The pixels in the first image whose pixel value is not equal to the white pixel value are taken as the first pixel, and the first pixels are aggregated to form the first element region; Each of the different pixel values ​​contained in the first element region is taken as the first pixel value. The number of pixels in the first element region whose pixel value is equal to the first pixel value is counted and taken as the first number. Divide the first number by the total number of pixels in the first element region to obtain the proportion of the first pixel value. Use the proportion of the first pixel value and the first pixel value as the distribution of the first pixel value in the first element region.

7. A high-speed visual inspection method for surface defects of microelectronic components according to claim 6, characterized in that, The process of obtaining the second element region and analyzing the distribution of the second pixel values ​​in the second element region includes the following steps: In the second image, pixels whose pixel values ​​are not equal to white pixel values ​​are taken as second pixels, and these second pixels are aggregated to form a second element region. Each of the different pixel values ​​contained in the second element region is taken as the second pixel value. The number of pixels in the second element region whose pixel value is equal to the second pixel value is counted and taken as the second number. Divide the second number by the total number of pixels in the second element region to obtain the proportion of the second pixel value. Use the proportion of the second pixel value and the second pixel value as the distribution of the second pixel value in the second element region.

8. A high-speed visual inspection method for surface defects of microelectronic components according to claim 7, characterized in that, The process of selecting the first target image from the first image includes the following steps: The first image with the smallest number of first pixel values ​​is selected as the first target image.

9. A high-speed visual inspection method for surface defects of microelectronic components according to claim 8, characterized in that, The process of selecting the second target image from the second image includes the following steps: Establish a correspondence between the second pixel value in the second image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the second pixel value in the second image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the second image relative to the sample stitched image. The minimum error value of the second image relative to at least one sample stitched image is taken as the actual error ratio of the second image, and the second image with the smallest actual error ratio is taken as the second target image.

10. A high-speed visual inspection method for surface defects of microelectronic components according to claim 9, characterized in that, The process of determining whether an electronic component has a defect includes the following steps: Establish a correspondence between the first pixel value in the first target image and the equal sample pixel value in the sample stitched image; The absolute value of the difference between the proportion of the first pixel value in the first target image and the proportion of the corresponding sample pixel value in the sample stitched image is accumulated to obtain the error value of the first target image relative to the sample stitched image. The minimum error value of the first target image relative to at least one sample stitched image is taken as the actual error ratio of the first target image; If the actual error ratio of the first target image and the actual error ratio of the second target image both do not exceed the allowable error ratio, then the electronic component is not defective; otherwise, the electronic component is defective.

Images