Anti-damage feeding system and method for chip braiding machine based on computer vision

Abstract

The invention discloses an anti-damage feeding system and method for a chip taping machine based on computer vision, and relates to the technical field of automatic control. The feeding method comprises the steps that based on deployment of a fixed-focus industrial camera module and a three-point matrix light source, before a chip taping machine carries out feeding propelling operation on a chip, the fixed-focus industrial camera module and the three-point matrix light source are arranged; the chips before being pushed in the feeding track are shot, and an original image data matrix of the chips before being pushed is formed; in combination with a second order gradient algorithm, judging whether the image acquisition effect of the chip before propulsion is suitable for subsequent chip posture recognition under the current lighting condition, and sending out a corresponding posture recognition instruction; according to the eccentricity degree and the relative space inclination angle state of the chip before being propelled and in combination with the image quality coefficient, whether the posture of the chip before being propelled causes damage to the chip in the feeding and propelling process of the chip taping machine or not is judged, and a feeding control instruction of a corresponding grade is sent out; and executing a corresponding anti-damage feeding propulsion action to reduce the damage to the chip.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation control, in particular to a computer vision-based anti-damage feeding system and method for a chip taping machine. BACKGROUND

[0002] In the current integrated circuit chip packaging and delivery process, as a key device for realizing chip automatic taping and packaging, the chip taping machine needs to ensure that each chip is in a relatively accurate and undamaged posture before feeding to avoid misplacement, skewing or corner damage of the chip during the feeding process. Therefore, it is urgent to introduce an efficient and stable visual image acquisition and posture judgment mechanism in the feeding link to realize high-precision identification of the chip state before pushing and feeding control, and to improve the reliability and safety of the overall automatic assembly.

[0003] In the existing feeding method of the chip taping machine, although an industrial camera is generally used for image acquisition, the image is acquired with fixed exposure parameters, which does not have the dynamic judgment ability of image clarity and uniformity of illumination, resulting in the problem of reflection and uneven brightness of the acquired image under different lighting conditions, which seriously affects the identification accuracy of the subsequent chip posture. In addition, the posture judgment in the existing scheme only relies on edge geometry and template matching, and lacks a comprehensive modeling and evaluation mechanism for the local space angle and eccentric distance in the chip image, making it difficult to accurately determine whether the chip is in a deflection and displacement state, thereby making it difficult to provide reliable basis for the feeding advance action. Therefore, there is an urgent need for a feeding pre-judgment method that combines image quality evaluation and multi-dimensional posture parameter extraction to effectively solve the problem of chip damage caused by poor image quality and insufficient posture shift judgment in the feeding process. SUMMARY

[0004] To overcome the deficiencies of the prior art, the present application provides a computer vision-based anti-damage feeding system and method for a chip taping machine, which solves the problems in the background art.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a computer vision-based anti-damage feeding method for a chip taping machine, comprising the following steps: S1, based on the deployment of a fixed-focus industrial camera module and a three-point matrix light source, before the chip taping machine performs the feeding and advancing operation on the chip, the chip before advancing in the feeding track is photographed to form an original image data matrix of the chip before advancing; S2, based on the original image data matrix, combined with a second-order gradient algorithm, determine the image quality coefficient of the chip before advancing under the current lighting condition to judge whether the image acquisition effect of the chip before advancing under the current lighting condition is suitable for subsequent chip posture recognition, and issue corresponding posture recognition instructions; S3. After receiving the attitude analysis signal, based on the original image data matrix, determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it was pushed forward. Combined with the image quality coefficient, analyze the degree of attitude deviation of the chip before it was pushed forward, so as to determine the attitude deviation index of the chip before it was pushed forward. S4. Based on the chip's attitude deviation index before it is pushed forward, determine whether the chip's current attitude before it is pushed forward will cause damage to the chip during the chip tape and reel feeding process, issue the corresponding level of feeding control command, and execute the corresponding anti-damage feeding and pushing action.

[0006] Preferably, step S1 specifically includes: S11. Deploy a fixed-focus industrial camera module and a three-point matrix light source on the feeding track plane of the chip taping machine. The specific deployment process is as follows: The working field of view and image projection area of ​​the industrial camera module are determined based on the feeding track plane of the chip taping machine. The optical axis of the industrial camera module is vertically aligned with the feeding track plane of the chip taping machine. At the same time, a three-point matrix light source is deployed directly above the feeding track surface of the chip taping machine. The three-point matrix light source includes three sets of LED arrays, which are located at the upper left, upper right and upper front, and illuminate the surface at an oblique angle. S12. Before the chip feeding and pushing operation of the chip tape and reel machine, the chips in the feeding track before they are pushed forward are photographed to form the original image data matrix of the chips before they are pushed forward. The specific process is as follows: During continuous illumination by the three-point matrix light source, the image acquisition controller of the industrial camera module, based on the feeding status signal of the chip tape machine, fixes the image acquisition time of the industrial camera module within the static time window of the chip before it is advanced. The industrial camera module then performs a complete exposure and readout operation within this static time window. After grayscale processing, the original image data matrix of the chip before it is advanced is formed. The original image data matrix Ghd... xy This indicates that the grayscale value of the corresponding pixel unit in the plane coordinates (x, y) is Ghd. i ; S13. Perform preprocessing operations on the original image data matrix of the chip before it is pushed forward. The preprocessing operations include mean filtering and contrast normalization.

[0007] Preferably, step S2 specifically includes: S21. Based on the original image data matrix of the chip before advancement, determine the reflection interference coefficient of each pixel unit. The specific process is as follows: Feature recognition is performed on the original image data matrix of the chip before it is pushed forward in S12. The gray value of each pixel unit of the chip before it is pushed forward in the plane coordinate is extracted. Combined with the statistical mean algorithm, the average gray value of the pixels in the plane coordinate is obtained. The average grayscale value of each pixel in the planar coordinate system is correlated with the grayscale value of each pixel unit in the corresponding planar coordinate system. Based on the brightness response difference model under multi-angle light source illumination, and after dimensionless processing, the degree of reflective interference of each pixel unit in the planar coordinate system is analyzed to determine the reflective interference coefficient of each pixel unit. Specifically: In the formula, Xgr i This represents the reflection interference coefficient of the corresponding pixel unit, in Ghb. i This represents the grayscale value of the corresponding pixel unit. Represents the average grayscale value of pixels in a planar coordinate system. This represents the cosine value of the angle between the normal axis of the corresponding pixel unit and the direction of the incident light.

[0008] Preferably, in step S22, the original image data matrix of the chip before advancement is correlated with the reflection interference coefficient of each pixel unit in S21. Using a second-order gradient algorithm, the overall image clarity and anti-reflection level of the chip before advancement under the current lighting conditions are analyzed to determine the image quality coefficient of the chip before advancement under the current lighting conditions. Specifically: In the formula, Ghd represents the image quality coefficient of the chip before it is advanced under the current lighting conditions. xy Represents the original image data matrix. This represents the sum of the second-order grayscale derivatives of the corresponding pixel unit in the horizontal and vertical directions. Xgr represents the partial derivative sign. i This represents the reflection interference coefficient of the corresponding pixel unit, where n represents the number of pixel units.

[0009] Preferably, in step S23, the image acquisition effect of the chip before advancement under the current lighting conditions is analyzed to determine whether it is suitable for subsequent chip attitude recognition, and corresponding attitude recognition commands are issued. The specific analysis process is as follows: If the image quality coefficient of the chip before advancement under the current lighting conditions exceeds the preset quality threshold, it indicates that the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and an attitude analysis signal is issued. If the image quality coefficient of the chip before advancement does not exceed the preset quality threshold under the current lighting conditions, it means that the image acquisition effect of the chip before advancement under the current lighting conditions is not suitable for subsequent chip attitude recognition. At this time, the intensity of the three-point matrix light source and the exposure time of the industrial camera module are adjusted, and the image acquisition operation of the chip before advancement is repeated.

[0010] Preferably, step S3 specifically includes: S31. After receiving the attitude analysis signal, based on the original image data matrix of the chip before advancement, edge detection and closed contour extraction methods are used to identify the two-dimensional outer contour of the chip before advancement to determine the coordinates of the image center point. Based on the standard size parameters of the chip taping machine's feeding track plane, the coordinates of the track center point of the chip taping machine's feeding track plane are extracted. After calculating the Euclidean distance, the eccentricity distance of the chip before advancement is determined, specifically: In the formula, Dpx represents the eccentric distance of the chip before advancement, (x1-x2) represents the distance between the center of the track and the center of the image on the x-axis, (y1-y2) represents the distance between the center of the track and the center of the image on the y-axis, and R represents the scaling factor.

[0011] Preferably, in step S32, pixel-level feature matching is performed on the original image data matrix to extract the relative spatial tilt angle data matrix of the chip before advancement. The specific extraction process is as follows: By acquiring the left and right views from the original image data matrix separately and performing pixel-level feature matching between the view images, the position offset values ​​of the corresponding pixel units are extracted to construct a three-dimensional depth map of the chip before it is advanced. Extract the maximum gradient line segment from the high gradient region of the 3D depth map, and extract the angle between the maximum gradient line segment and the reference horizontal plane of the chip tape machine's feeding track. Obtain the relative spatial tilt angle data matrix of the chip before advancement, specifically in the form: Tα=[Tα x ,Tα y ,Tα z ]; In the formula, Tα x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y Tα represents the tilt angle of the chip in the y-axis direction before it is advanced. z This indicates the tilt angle of the chip in the z-axis direction before it is advanced.

[0012] Preferably, in step S33, the relative spatial tilt angle data matrix of the chip before advancement is correlated with the eccentricity distance, and combined with the image quality coefficient of the chip before advancement under the current lighting conditions, after dimensionless processing, the attitude deviation degree of the chip before advancement under the current lighting conditions is analyzed to determine the attitude deviation index of the chip before advancement under the current lighting conditions. Specifically: In the formula, Zzt represents the feeding posture offset index of the chip tape-and-reel machine under the current illumination conditions, and Tα x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y represents the tilt angle of the chip in the y-axis direction before advancement, Dpx represents the eccentricity distance of the chip before advancement, Xtz represents the image quality coefficient of the chip before advancement, and arctan(*) represents the tangent function.

[0013] Preferably, step S4 specifically includes: S41. Based on the obtained attitude deviation index of the chip before advancement under the current lighting conditions, compare it with a preset deviation threshold to determine whether the chip's attitude before advancement under the current lighting conditions will cause damage to the chip during the chip tape and reel feeding process. Issue a corresponding level of feeding control command and execute the corresponding anti-damage feeding and advancing action, specifically: If the attitude deviation index of the chip before advancement under the current lighting conditions is greater than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a first-level feeding control command is issued. The chip tape and reel machine stops feeding and advancing operation and prompts the operator to reposition the chip. If the attitude deviation index of the chip before advancement under the current lighting conditions is equal to the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a secondary feeding control command is issued, and the chip tape and reel machine performs a flexible feeding and advancing operation. If the attitude deviation index of the chip before advancement under the current lighting conditions is less than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a level 3 feeding control command is issued, and the chip tape and reel machine performs normal feeding and advancing operation.

[0014] A computer vision-based chip taping machine anti-damage feeding system includes a vision acquisition module, an image quality analysis module, an attitude analysis module, and a feeding control module. The vision acquisition module is based on a fixed-focus industrial camera module and a three-point matrix light source. Before the chip feeding and pushing operation of the chip tape machine, it takes pictures of the chip in the feeding track before it is pushed, forming the original image data matrix of the chip before it is pushed. The image quality analysis module, based on the original image data matrix and combined with the second-order gradient algorithm, determines the image quality coefficient of the chip before advancement under the current lighting conditions. This is used to determine whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and to issue corresponding attitude recognition commands. The attitude analysis module is used to determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it is pushed forward, based on the original image data matrix after receiving the attitude analysis signal. Combined with the image quality coefficient, it analyzes the degree of attitude deviation of the chip before it is pushed forward, so as to determine the attitude deviation index of the chip before it is pushed forward. The feeding control module is used to determine whether the current attitude of the chip before it is pushed will cause damage to the chip during the feeding process of the chip tape machine, based on the attitude deviation index of the chip before it is pushed. It then issues a corresponding level of feeding control command and executes the corresponding anti-damage feeding action.

[0015] This invention provides a computer vision-based chip taping machine anti-damage feeding system and method, which has the following beneficial effects: (1) By constructing a progressive recognition and control process with “image quality assessment - posture parameter extraction - risk judgment - feeding action response” as the main line, the chip damage caused by unstable image acquisition effect, single posture judgment basis and blind feeding action in the existing chip taping process is systematically solved; by multi-source lighting, image grayscale matrix modeling and reflection interference analysis, the acquisition stability of image data is improved; at the same time, the posture offset index is constructed by combining posture tilt angle, eccentric distance and image quality parameters to realize quantitative prediction and dynamic decision-making of potential feeding risks, thereby ensuring the safety and accuracy of chip feeding process and significantly improving the automation level and yield of chip taping machine feeding link.

[0016] (2) By adopting a fixed-focus industrial camera module and a three-point matrix light source in a coordinated deployment manner, combined with a fixed exposure window and a dynamic illumination compensation mechanism, the image clarity is effectively reduced due to uneven illumination and reflection interference in the existing system. By constructing an original image data matrix, supplemented by pixel grayscale mean comparison and second-order gradient edge change detection, the image quality coefficient is output in a comprehensive manner, which can determine the clarity and stability of the currently acquired image in real time, ensuring that the acquired image has high edge sharpness and low reflection disturbance level. This mechanism not only improves the distinguishability of chip edges and geometric features, but also provides a stable and reliable input data source for subsequent posture feature extraction and posture judgment algorithms.

[0017] (3) Based on the image acquisition quality assessment results, under the premise that the image quality is qualified, the chip's eccentricity distance and relative spatial tilt angle data are further extracted, and the attitude deviation index is constructed as the core criterion for measuring the attitude risk before chip feeding; by performing multi-parameter fusion calculation on the degree of deviation between the image center and the track center, the three-dimensional tilt angle and the angle between the track reference plane, and combining the image quality coefficient under the current lighting conditions, a high-precision analysis of the chip's current spatial attitude state is completed; on this basis, a feeding risk level judgment and damage prevention control command are formed, thereby realizing a multi-level control response mechanism of flexible feeding, prediction correction and automatic pause intervention, significantly improving the safety and precision control capability of chip propulsion action, and avoiding chip corner collisions and structural damage caused by attitude deviation. Attached Figure Description

[0018] Figure 1This is a schematic diagram of a chip taping machine anti-damage feeding method based on computer vision according to the present invention. Figure 2 This is a block diagram of a chip taping machine anti-damage feeding system based on computer vision according to the present invention; Figure 3 This is a logic diagram of parts S1 and S2 of the present invention; Figure 4 This is a logic diagram for parts S3 and S4 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 Please see Figure 1 This invention provides a computer vision-based chip taping machine anti-damage feeding method, comprising the following steps: S1. Based on the deployment of a fixed-focus industrial camera module and a three-point matrix light source, before the chip feeding and pushing operation of the chip tape machine, the chip in the feeding track before being pushed is photographed to form the original image data matrix of the chip before being pushed. S2. Based on the original image data matrix and combined with the second-order gradient algorithm, determine the image quality coefficient of the chip before advancement under the current lighting conditions, so as to determine whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and issue the corresponding attitude recognition command. S3. After receiving the attitude analysis signal, based on the original image data matrix, determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it was pushed forward. Combined with the image quality coefficient, analyze the degree of attitude deviation of the chip before it was pushed forward, so as to determine the attitude deviation index of the chip before it was pushed forward. S4. Based on the chip's attitude deviation index before it is pushed forward, determine whether the chip's current attitude before it is pushed forward will cause damage to the chip during the chip tape and reel feeding process, issue the corresponding level of feeding control command, and execute the corresponding anti-damage feeding and pushing action.

[0021] In this embodiment, by constructing a closed-loop process of "image acquisition—quality assessment—pose recognition—risk judgment—feeding control," high-precision dynamic recognition and flexible damage prevention control of the chip pre-feeding state are achieved, significantly improving the safety and intelligence level of the automated feeding process of the chip taping machine. The significant advantages of this method are: firstly, by combining a fixed-focus industrial camera module with a multi-angle three-point matrix light source, high-quality raw image data can be acquired during the chip's stationary phase, avoiding reflection interference and image blurring problems caused by traditional fixed light sources; secondly, by introducing a second-order gradient algorithm to perform high-order analysis of image grayscale changes and constructing image quality coefficients, it is possible to judge the image quality immediately after image acquisition. The image input data is analyzed for clarity and edge saliency to ensure a good foundation for recognition. Furthermore, the chip's eccentricity distance and three-dimensional tilt angle parameters are extracted from the original image data matrix to construct an attitude offset index, effectively overcoming the limitation of traditional edge fitting methods in quantifying chip offset states. Finally, control commands are generated hierarchically based on the attitude offset index, achieving multi-level feeding action control from automatic pause to flexible advancement. This ensures the integrity of the chip structure while also providing higher adaptive processing capabilities. Therefore, this method not only has outstanding effects in improving chip feeding accuracy but also demonstrates unique application value in enhancing the intelligent sensing capabilities and reliability of the overall tape and reel equipment.

[0022] Example 2 Please refer to Figure 1 and Figure 3 Specifically: S1 includes the following steps: S11. Deploy a fixed-focus industrial camera module and a three-point matrix light source on the feeding track plane of the chip taping machine. The specific deployment process is as follows: The working field of view and image projection area of ​​the industrial camera module are determined based on the feeding track plane of the chip taping machine. The optical axis of the industrial camera module is vertically aligned with the feeding track plane of the chip taping machine. At the same time, a three-point matrix light source is deployed directly above the feeding track surface of the chip taping machine. The three-point matrix light source includes three sets of LED arrays, which are located at the upper left, upper right and upper front, and illuminate the surface at an oblique angle. The three-point matrix light source refers to three arrayed LED light sources deployed above the surface of the chip taping machine's feed track, located at the upper left, upper right, and upper front of the chip, respectively, illuminating the chip feed track at an oblique incidence angle. Its main function is to effectively enhance the light and shadow contrast between the chip edge and surface structure by obliquely projecting light from multiple angles, thereby reducing problems such as shadow dead angles, reflection interference, and uneven brightness caused by a single light source during the imaging process. This lighting method helps to improve the clarity of the chip outline and the overall illumination balance in the image, providing a more stable and high-quality visual data foundation for subsequent image quality analysis, edge recognition, and posture judgment. It is a key lighting structure design for achieving highly robust image acquisition. S12. Before the chip feeding and pushing operation of the chip tape and reel machine, the chips in the feeding track before they are pushed forward are photographed to form the original image data matrix of the chips before they are pushed forward. The specific process is as follows: During continuous illumination by the three-point matrix light source, the image acquisition controller of the industrial camera module, based on the feeding status signal of the chip tape machine, fixes the image acquisition time of the industrial camera module within the static time window of the chip before it is advanced. The industrial camera module then performs a complete exposure and readout operation within this static time window. After grayscale processing, the original image data matrix of the chip before it is advanced is formed. The original image data matrix Ghd... xy This indicates that the grayscale value of the corresponding pixel unit in the plane coordinates (x, y) is Ghd. i ; The raw image data matrix refers to a set of pixel grayscale values ​​represented by two-dimensional coordinates (x, y) after the chip is photographed in a static state by an industrial camera module before the chip advancement operation, and then processed to grayscale. The grayscale value of each pixel unit is Ghd. i The original image data matrix effectively reflects the visible light reflection characteristics of the chip on the feeding track plane, and serves as the original input basis for subsequent image quality analysis, posture recognition, and offset judgment calculation modules. The acquisition process of the original image data matrix relies on the uniform illumination provided by a three-point matrix light source, which, together with the industrial camera module, performs synchronous exposure and data reading within the chip's stationary time window. This ensures that the acquired images are both clear and have complete illumination information, effectively depicting the chip's boundary and relative position features, and providing core data support for achieving high-precision visual recognition and damage prevention control. S13. Perform preprocessing operations on the original image data matrix of the chip before it is pushed forward. The preprocessing operations include mean filtering and contrast normalization.

[0023] In this embodiment, by deploying a high-precision fixed-focus industrial camera module and a three-point matrix light source, combined with a synchronous exposure mechanism within a static time window, stable and high-fidelity image acquisition of the chip state before advancement is achieved, improving the stability of image input and structural restoration capability. The unique advantage of this process lies in: by vertically positioning the optical axis of the industrial camera module on the feeding track plane, it ensures that the captured image has minimal geometric distortion and maximum structural alignment, providing an ideal basis for subsequent attitude analysis and error estimation; simultaneously, the oblique incidence configuration of the three-point matrix light source not only effectively suppresses the hard-edge shadows formed under traditional single light sources, but also generates rich brightness gradient features on the surface of complex packaging materials, thereby enhancing the key structures of boundaries and corners in the image. The system exhibits high responsiveness; during image acquisition, it precisely locks the static time window based on the status signal in the chip feeding trajectory, performing complete exposure and readout within this time window to ensure temporal consistency and signal integrity of the image data; further, through grayscale processing and contrast normalization preprocessing steps, it not only improves the robustness of image data under complex lighting conditions but also enhances the adaptability of subsequent image gradient and edge extraction algorithms; in summary, this acquisition mechanism not only solves the problems of reflection interference, exposure deviation, and geometric distortion in traditional image acquisition methods but also lays a highly consistent and feature-clear foundation of input conditions for image quality assessment, attitude analysis, and dynamic decision-making in subsequent modules, serving as a key starting point for ensuring the effective operation of the damage-resistant feeding control system.

[0024] Example 3 Please refer to Figure 1 and Figure 3 Specifically: The specific steps of S2 include: S21. Based on the original image data matrix of the chip before advancement, determine the reflection interference coefficient of each pixel unit. The specific process is as follows: Feature recognition is performed on the original image data matrix of the chip before it is pushed forward in S12. The gray value of each pixel unit of the chip before it is pushed forward in the plane coordinate is extracted. Combined with the statistical mean algorithm, the average gray value of the pixels in the plane coordinate is obtained. The average grayscale value of each pixel in the planar coordinate system is correlated with the grayscale value of each pixel unit in the corresponding planar coordinate system. Based on the brightness response difference model under multi-angle light source illumination, and after dimensionless processing, the degree of reflective interference of each pixel unit in the planar coordinate system is analyzed to determine the reflective interference coefficient of each pixel unit. Specifically: In the formula, Xgr i This represents the reflection interference coefficient of the corresponding pixel unit, in Ghb. i This represents the grayscale value of the corresponding pixel unit. Represents the average grayscale value of pixels in a planar coordinate system. This represents the cosine value of the angle between the normal axis of the corresponding pixel unit and the direction of the incident light.

[0025] It should be noted that the formula in S21 is used to quantify the degree of grayscale disturbance caused by local reflection in each pixel unit of the image, which is one of the important parameters for identifying the quality of image acquisition. This formula comprehensively considers two key factors: first, the relative deviation of the current pixel grayscale value from the overall image grayscale mean; and second, the cosine of the angle between the normal axis of the pixel unit and the incident light direction, i.e., the geometric intensity relationship of the illumination it receives. After dimensionless normalization, a reflection interference coefficient matrix is ​​formed, reflecting whether there are non-uniform phenomena such as bright spots, overexposure, or sudden changes in brightness in the local area of ​​the image under the current lighting conditions. This process is directly related to the existing chip tape-and-reel machine feeding method, as pointed out in the background technology, which is prone to problems under fixed exposure conditions. Image reflection and uneven brightness lead to attitude analysis failure. The reflection interference modeling mechanism established by the formula is designed to address the technical pain point of lacking dynamic quantitative indicators for image sharpness discrimination. It can not only be used to construct subsequent image quality coefficients and jointly determine the effectiveness of image acquisition with the second-order gradient derivative index, but also serve as a basis for feedback adjustment of lighting parameters and exposure time, thereby achieving closed-loop optimization. By meticulously characterizing the reflection intensity of each pixel, this mechanism effectively improves the sensitivity and resolution of image quality assessment, providing a high-precision basis for whether the chip has reliable attitude recognition conditions, further ensuring that the chip can be stably identified before feeding and preventing misalignment and damage, and realizing the image-driven intelligent upgrade of tape and reel operation.

[0026] Specifically, in S22, the original image data matrix of the chip before advancement is correlated with the reflection interference coefficient of each pixel unit in S21. Using a second-order gradient algorithm, the overall image clarity and anti-reflection level of the chip before advancement under the current lighting conditions are analyzed to determine the image quality coefficient of the chip before advancement under the current lighting conditions. In the formula, Ghd represents the image quality coefficient of the chip before it is advanced under the current lighting conditions. xy Represents the original image data matrix. This represents the sum of the second-order grayscale derivatives of the corresponding pixel unit in the horizontal and vertical directions. Xgr represents the partial derivative sign. i This represents the reflection interference coefficient of the corresponding pixel unit, i = 1, 2, 3, ..., n, where n represents the number of pixel units.

[0027] It should be noted that the formula in S22 is used to quantitatively evaluate the image sharpness and anti-reflective interference level before the chip is advanced, under the current lighting conditions, providing a key criterion for whether to perform attitude recognition subsequently. This formula uses the second-order gradient change of the original image data matrix—that is, the sum of the second derivatives of the image grayscale in the horizontal and vertical directions—as a factor reflecting the overall image sharpness. Simultaneously, it incorporates the reflective interference coefficient of each pixel, introducing it in the denominator to suppress local image distortion caused by reflection, thus reflecting the image's anti-reflective capability. The introduction of the second-order gradient solves the problem of dynamically evaluating image sharpness mentioned in the background technology, while the linkage control of the reflective interference coefficient... This design precisely addresses the technical shortcomings of uneven image brightness and severe reflection interference, thus forming a comprehensive image quality evaluation model that balances grayscale detail sharpness and illumination consistency. The actual function of the image quality coefficient is to determine whether the image meets the prerequisites for pose feature extraction and judgment after the system captures the image. If the image quality is too low, the lighting and exposure parameters will be automatically adjusted and the image will be re-acquired, effectively avoiding the risk of misjudgment caused by image blur or illumination imbalance. Therefore, this formula forms a close logical loop with the requirements in the background technology, ensuring that the chip has sufficient basic image quality before visual recognition, thereby improving the recognition stability and material supply security of the chip tape and reel system from the root.

[0028] Specifically, S23 analyzes whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and issues corresponding attitude recognition commands. The specific analysis process is as follows: If the image quality coefficient of the chip before advancement under the current lighting conditions exceeds the preset quality threshold, it indicates that the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and an attitude analysis signal is issued. If the image quality coefficient of the chip before advancement does not exceed the preset quality threshold under the current lighting conditions, it means that the image acquisition effect of the chip before advancement under the current lighting conditions is not suitable for subsequent chip attitude recognition. At this time, the intensity of the three-point matrix light source and the exposure time of the industrial camera module are adjusted, and the image acquisition operation of the chip before advancement is repeated.

[0029] It should be noted that adjusting the intensity of the three-point matrix light source and the exposure time of the industrial camera module, and then re-performing the image acquisition operation before chip advancement, involves the following steps: prioritizing the reduction of the luminous intensity of each LED array in the three-point matrix light source, especially the incident direction with significant reflective interference; adjusting the exposure time parameters of the industrial camera module, including exposure duration and gain value, to achieve an optimized balance between brightness and sharpness; this process is usually automatically executed by the image acquisition control unit, and iteratively corrected based on real-time feedback of the image quality coefficient until the new image quality coefficient exceeds a preset threshold, ultimately completing an image acquisition operation that meets the posture recognition conditions; this mechanism significantly improves the system's adaptability to image quality in complex lighting environments.

[0030] It should be noted that the quality threshold is a reference value set during chip image acquisition and processing to determine whether an image meets the requirements for pose recognition, based on the image quality coefficient. Its function is to quickly assess whether the current image meets the minimum standards required for pose recognition in terms of sharpness, contrast, and anti-reflective interference, thereby deciding whether to proceed to the subsequent pose analysis process. The preset quality threshold is determined by testing and statistical analysis of a large number of sample images, combined with the recognition effect of images under different lighting and exposure conditions in actual use environments, using the relationship curve between the image quality coefficient and the recognition success rate to determine the critical point as a reference standard.

[0031] In this embodiment, by introducing a reflection interference coefficient modeling and image quality coefficient quantization mechanism, the evaluation accuracy and dynamic response capability of the chip's image acquisition effect are improved, laying a solid foundation for the stability and accuracy of subsequent posture recognition. Its technical advantage lies in that it no longer relies solely on image brightness or edge sharpness as a single criterion, but instead integrates the statistical characteristics of grayscale distribution and the spatial incident light interaction response to construct the reflection interference coefficient of each pixel unit under real physical lighting conditions. This parameter is modeled by combining the difference between pixel grayscale values ​​and local average brightness in the original image data matrix with the cosine function of the angle between the pixel normal axis and the incident light direction, which can truly reflect the degree of reflection influence on the pixel area, thereby avoiding image misjudgment problems caused by differences in material reflection characteristics. Based on this, the grayscale of each pixel in the image is further extracted using a second-order gradient algorithm. The system analyzes the changing trends, constructs the derivatives of grayscale response changes in the horizontal and vertical directions of the image, and performs normalized accumulation processing in conjunction with reflection interference factors. Finally, it outputs a unified image quality coefficient index, thereby achieving a comprehensive evaluation of the image's clarity, contrast, and anti-reflection capability under the current lighting configuration. More importantly, the image quality coefficient serves as a dynamic threshold, triggering the entry into the posture recognition process. If the recognition threshold is not met, the system automatically adjusts the light source intensity and industrial camera exposure parameters, and re-acquires the image, ensuring that the images used for posture analysis are in optimal lighting and edge recognition conditions. This image evaluation method, which integrates reflection modeling and multi-order gradient response, not only improves the physical rationality of the judgment but also endows the system with adaptability and closed-loop control capabilities under actual industrial conditions. This is a key prerequisite for achieving stable operation of the entire damage-prevention material supply process.

[0032] Example 4 Please refer to Figure 1 and Figure 4 Specifically: The specific steps of S3 include: S31. After receiving the attitude analysis signal, based on the original image data matrix of the chip before advancement, edge detection and closed contour extraction methods are used to identify the two-dimensional outer contour of the chip before advancement to determine the coordinates of the image center point. Based on the standard size parameters of the chip taping machine's feeding track plane, the coordinates of the track center point of the chip taping machine's feeding track plane are extracted. After calculating the Euclidean distance, the eccentricity distance of the chip before advancement is determined, specifically: In the formula, Dpx represents the eccentric distance of the chip before advancement, (x1-x2) represents the distance between the center of the track and the center of the image on the x-axis, (y1-y2) represents the distance between the center of the track and the center of the image on the y-axis, and R represents the scaling factor.

[0033] It should be noted that the formula in S31 is used to calculate the eccentricity distance of the chip before it is advanced, that is, the degree of relative displacement between the chip image center and the chip taping machine feeding track center. Specifically, it is based on the geometric distance difference between the x-axis and y-axis on the two-dimensional coordinate plane of the image, calculated using the Euclidean formula, and multiplied by a scaling factor R to reflect the degree of offset on the actual physical scale. The design of this formula directly corresponds to the problem in the background technology that the chip may be skewed or displaced before feeding, but it is difficult to accurately determine this. The background technology points out that the existing solution lacks the ability to determine whether the chip position is at the center of the feeding trajectory, and this eccentricity distance formula is designed to fill this technical gap. By using the coordinate difference between the image center and the feeding trajectory center, it provides a quantitative basis for whether the chip is on a compliant feeding path. Its role is not only to provide a judgment on whether it is aligned, but also to provide key input parameters for subsequent attitude offset index analysis, thereby further linking the judgment on whether the chip advancement action is safe. Therefore, eccentricity distance is one of the core indicators in image geometric stability analysis. Its introduction improves the positioning accuracy before feeding, the accuracy of feeding decision, and the reliability of chip damage prevention control capabilities. Specifically, S32, pixel-level feature matching is performed on the original image data matrix to extract the relative spatial tilt angle data matrix of the chip before advancement. The specific extraction process is as follows: By acquiring the left and right views from the original image data matrix separately and performing pixel-level feature matching between the view images, the position offset values ​​of the corresponding pixel units are extracted to construct a three-dimensional depth map of the chip before it is advanced. In constructing a 3D depth map of the chip before its advancement, an industrial camera module acquires left and right view images at the same time from slightly offset positions on both sides. These two images record the 2D grayscale distribution of the chip from different perspectives. Based on image registration and pixel-level feature matching algorithms, the grayscale features, edge structures, or corner positions of key areas are compared pixel by pixel between the left and right views to extract the horizontal displacement between each pair of matching pixels. This data, combined with the known camera baseline distance and lens parameters, is used to calculate the spatial depth value of each point on the chip surface according to the principle of triangulation, thereby generating a 3D depth map containing complete spatial coordinate information. This provides an accurate spatial data foundation for subsequent attitude tilt and eccentricity analysis. This method can effectively capture subtle tilts and height differences on the chip surface and is a key step in improving the accuracy of attitude judgment. Extract the maximum gradient line segment from the high gradient region of the 3D depth map, and extract the angle between the maximum gradient line segment and the reference horizontal plane of the chip tape machine's feeding track. Obtain the relative spatial tilt angle data matrix of the chip before advancement, specifically in the form: Tα=[Tα x ,Tα y ,Tα zIn the formula, Tα represents the relative spatial tilt angle data matrix. x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y Tα represents the tilt angle of the chip in the y-axis direction before it is advanced. z This indicates the tilt angle of the chip in the z-axis direction before it is advanced.

[0034] It should be noted that the acquisition process of the aforementioned relative spatial tilt angle data matrix is ​​based on the identification and geometric fitting analysis of high gradient regions in the image space of the 3D depth map. After pixel-level feature matching is completed based on left and right view imaging, the 3D depth map of the chip before advancement is reconstructed through parallax calculation. Each pixel in the depth map reflects its relative height change in space. Gradient analysis is performed on this depth map to identify the high gradient regions with the fastest grayscale changes in different directions, and the maximum gradient direction line segment is extracted in these regions as the feature line with the most significant spatial tilt change on the chip surface. The angle between the spatial direction vector of the maximum gradient line segment and the standard horizontal plane direction vector of the chip taping machine feeding track is calculated. Based on the 3D vector angle calculation formula, the tilt angle of the chip relative to the horizontal reference plane of the track in the x, y, and z directions is obtained, denoted as Tα. x , Tα y and Tα z These tilt angles can reflect whether the chip is tilted, warped, or flipped. They are of great reference value for judging the stability and stress safety of the chip during the feeding process and are key parameters for achieving accurate three-dimensional attitude judgment and propulsion control decision-making.

[0035] Specifically, S33, based on the relative spatial tilt angle data matrix of the chip before advancement and its eccentricity distance, and combined with the image quality coefficient of the chip before advancement under the current lighting conditions, after dimensionless processing, the degree of attitude deviation of the chip before advancement under the current lighting conditions is analyzed to determine the attitude deviation index of the chip before advancement under the current lighting conditions, specifically: In the formula, Zzt represents the feeding posture offset index of the chip tape-and-reel machine under the current illumination conditions, and Tα x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y denoted by y, Dpx represents the eccentricity of the chip before advancement, Xtz represents the image quality coefficient of the chip before advancement, arctan(*) represents the tangent function, and log(*) represents the logarithmic function.

[0036] It should be noted that the formula in S33 is used to calculate the attitude offset index, a quantitative indicator for evaluating the attitude stability of the chip in the non-advancing state. It reflects the coupled offset risk among chip tilt, positional deviation, and image sharpness. This formula combines the relative spatial tilt angle of the chip in the x-axis and y-axis directions with the chip's eccentricity distance to characterize the attitude deviation risk of the chip on the two-dimensional feeding plane. Then, dimensionless normalization is performed using the current image quality coefficient as the denominator to make the index adaptable to changes in illumination and comparable. As mentioned in the background technology, existing methods are difficult to effectively identify the combined effect of local spatial tilt angle and chip center point offset in the image, leading to attitude misjudgment or recognition failure, which in turn causes damage to the chip's corners during the advancement process. This formula is designed to address the above defects, and can jointly measure the chip's rotation angle and positional deviation to form a standardized attitude offset index. The higher the value, the more significant the chip's deflection or offset risk. The core function of this index is to provide a reliable basis for subsequent feeding strategy decisions.

[0037] In this embodiment, by extending the attitude judgment process from traditional edge matching to multi-dimensional geometric feature modeling and angle space quantization, not only is precise identification of the chip's attitude state before advancement achieved, but the predictive accuracy and control response efficiency of the anti-damage feeding mechanism are also effectively improved. First, based on the original image data matrix, this step uses edge detection and closed contour extraction algorithms to accurately reconstruct the complete outer boundary of the chip in a two-dimensional view. The chip image center coordinates are determined by extracting the contour centroid, and then, combined with the structural calibration parameters of the feeding track, the coordinates of the track center point are obtained. Thus, the eccentricity distance is constructed using the Euclidean distance algorithm, providing a clear displacement parameter for the initial modeling of attitude offset. Second, pixel-level feature matching is further performed through dual-view acquisition to construct a three-dimensional depth map of the chip before advancement. The maximum gradient line segment in the principal axis tilt direction is extracted from the high gradient region of the depth map, and the tilt is extracted in conjunction with the feeding track reference plane. The tilt angles are quantified into relative spatial tilt angle data matrices in three directions, allowing non-rigid displacement factors such as chip rotation and tilt to be included in the analysis framework. Finally, the tilt angle parameters, eccentricity distance, and image quality coefficient are fused dimensionlessly to construct an attitude offset index. The nonlinear mapping relationship is modeled using tangent and inverse tangent functions, effectively improving the sensitivity and threshold adjustability of attitude state classification. Through this multi-parameter fusion index construction mechanism, accurate identification of subtle chip tilt and poor alignment is achieved without relying on template matching or traditional hard threshold judgment. This attitude recognition strategy, which highly integrates spatial geometric features and visual quality features, not only solves the problem of difficulty in quantifying the combined degree of tilt angle and displacement in existing technologies, but also provides a quantitative, stable, and physically meaningful decision basis for subsequent damage prevention feeding actions, significantly improving the system's adaptability and operational safety in complex chip feeding environments.

[0038] Example 5 Please refer to Figure 1 and Figure 4 Specifically: The specific steps of S4 include: S41. Based on the obtained attitude deviation index of the chip before advancement under the current lighting conditions, compare it with a preset deviation threshold to determine whether the chip's attitude before advancement under the current lighting conditions will cause damage to the chip during the chip tape and reel feeding process. Issue a corresponding level of feeding control command and execute the corresponding anti-damage feeding and advancing action, specifically: If the attitude deviation index of the chip before advancement under the current lighting conditions is greater than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a first-level feeding control command is issued. The chip tape and reel machine stops feeding and advancing operation and prompts the operator to reposition the chip. If the attitude deviation index of the chip before advancement under the current lighting conditions is equal to the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a secondary feeding control command is issued, and the chip tape and reel machine performs a flexible feeding and advancing operation. If the attitude deviation index of the chip before advancement under the current lighting conditions is less than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a level 3 feeding control command is issued, and the chip tape and reel machine performs normal feeding and advancing operation.

[0039] The offset threshold is a key criterion for determining whether there is a risk of damage to the chip's posture. Its acquisition process is based on a large amount of statistical data and experimental verification. Specifically, under different batches of chips and different combinations of tilt angles and eccentricities, various posture offset scenarios are simulated. The chip status after feeding is labeled and classified, and it is recorded whether there are any abnormalities such as edge cracking, jamming, or deviation during the feeding process. Subsequently, regression fitting is performed on the corresponding posture offset index and the actual damage incidence rate to form a response curve between posture offset and damage risk. A critical point is selected from this curve as the offset threshold to ensure that chips above the threshold have a high probability of damage, thereby implementing a pause or flexible feeding strategy based on the principle of safety and conservatism. In addition, this threshold can be parameterized and corrected in combination with chip packaging form, feeding structure characteristics, and visual image quality to enable it to have adaptive adjustment capabilities and improve its adaptability and accuracy in different working environments. The system employs three mechanisms: a pause-feed advance operation, which involves immediately halting the advance and prompting the operator via the human-machine interface to manually intervene and reposition the chip to completely avoid potential damage; a flexible advance operation, which involves reducing the advance speed and adjusting the push rod force and stepping rhythm to allow the chip to enter the tape gently, thereby minimizing the risk of breakage caused by micro-friction and stress concentration; and a normal advance operation, which involves accurately feeding the chip into the tape at a standard speed under a stable posture, ensuring both feed efficiency and chip integrity. These three mechanisms form a graded response closed-loop control strategy, effectively preventing chip jamming, breakage, or edge chipping caused by abnormal posture, significantly improving the overall stability and safety of the feed process.

[0040] In this embodiment, the material feeding damage prevention decision mechanism set in step S4, based on the quantitative evaluation results of the chip attitude offset index, constructs a hierarchical material feeding control logic, realizing the transformation from single-state recognition to multi-level response execution, effectively solving the problem that existing chip tape and reel machines have difficulty dynamically adjusting the intensity and rhythm of the pushing action during the material feeding process; this step forms a three-level response mode by comparing the attitude offset index with a preset offset threshold: when the offset threshold is exceeded, a first-level material feeding control command is immediately triggered, pausing the material feeding and prompting manual intervention, realizing real-time avoidance of potential damage risks; when it equals the offset threshold, a second-level material feeding control command is triggered, adopting a flexible pushing method to reduce the material feeding force and speed to ensure safe chip transition; and when it is below the offset threshold, a third-level response mode is activated. The hierarchical feeding control command executes routine feeding operations to maintain feeding efficiency. This dynamic hierarchical control strategy effectively reduces the probability of chip damage caused by abnormal posture while ensuring feeding continuity. It is particularly suitable for chip taping tasks with different batches, package sizes, or lighting conditions. More importantly, this mechanism is based on an offset index constructed by fusing preceding image quality coefficients and posture tilt parameters. Its judgment logic not only has a clear physical interpretation basis but also supports flexible adjustment of threshold parameters according to process requirements in practical applications, possessing high engineering adaptability and portability. Therefore, this step not only improves the intelligence level of the feeding control process but also significantly enhances the ability to perceive and respond to abnormal posture states, making it a key link in achieving precise damage-proof feeding control.

[0041] Specifically, examples of S3 and S4 calculations are as follows: The following data was obtained by processing images taken of chips before they were advanced in the feeding track using a three-point matrix light source: The coordinates of the image center are: (x1, y1) = (520, 340); The center coordinates of the tape track are: (x2, y2) = (500, 350); Image scaling factor: R = 0.05; Chip three-dimensional tilt angle: Tα x =4°, Tα y =3°; Image quality index: Xtz=0.78; Step 1: Calculate the eccentricity distance Dpx Using the Euclidean distance formula, substitute the data: Dpx = √(520−500) 2 +(340−350) 2 *0.05==1.11, √ represents the radical sign; Step 2: Calculate the attitude deviation index Zzt Convert the angle to radians and substitute the values; Tα x =4°=0.0698rad; Tα y =3°=0.0524rad; Calculate the sum of squares of the angles and take the tangent: arctan((0.0698) 2 +(0.0524) 2 )=arctan(0.00487+0.00275)=0.00762; Calculate the eccentricity: log(1+Dpx 2 )=log(1+1.118 2 ) = 0.352; Zzt is obtained by summarizing: Zzt=(0.00762+0.3520.78) / 0.78=0.461; If the preset offset threshold is 0.5, it is determined that the chip's posture before advancement under the current lighting conditions will not cause damage to the chip during the chip tape and reel feeding process, and a level 3 feeding control command is issued. At this time, the chip tape and reel machine performs normal feeding and advancement operation.

[0042] Example 6 Please refer to Figure 1 and Figure 2 Specifically: a chip taping machine anti-damage feeding system based on computer vision, including a vision acquisition module, an image quality analysis module, an attitude analysis module and a feeding control module; The vision acquisition module is based on a fixed-focus industrial camera module and a three-point matrix light source. Before the chip feeding and pushing operation of the chip tape machine, it takes pictures of the chip in the feeding track before it is pushed, forming the original image data matrix of the chip before it is pushed. The image quality analysis module, based on the original image data matrix and combined with the second-order gradient algorithm, determines the image quality coefficient of the chip before advancement under the current lighting conditions. This is used to determine whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and to issue corresponding attitude recognition commands. The attitude analysis module is used to determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it is pushed forward, based on the original image data matrix after receiving the attitude analysis signal. Combined with the image quality coefficient, it analyzes the degree of attitude deviation of the chip before it is pushed forward, so as to determine the attitude deviation index of the chip before it is pushed forward. The feeding control module is used to determine whether the current attitude of the chip before it is pushed will cause damage to the chip during the feeding process of the chip tape machine, based on the attitude deviation index of the chip before it is pushed. It then issues a corresponding level of feeding control command and executes the corresponding anti-damage feeding action.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A computer vision-based chip taping machine anti-damage feeding method, characterized in that: Includes the following steps: S1. Based on the deployment of a fixed-focus industrial camera module and a three-point matrix light source, before the chip feeding and pushing operation of the chip tape machine, the chip in the feeding track before being pushed is photographed to form the original image data matrix of the chip before being pushed. S2. Based on the original image data matrix and combined with the second-order gradient algorithm, determine the image quality coefficient of the chip before advancement under the current lighting conditions, so as to determine whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and issue the corresponding attitude recognition command. S3. After receiving the attitude analysis signal, based on the original image data matrix, determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it was pushed forward. Combined with the image quality coefficient, analyze the degree of attitude deviation of the chip before it was pushed forward, so as to determine the attitude deviation index of the chip before it was pushed forward. S4. Based on the chip's attitude deviation index before it is pushed forward, determine whether the chip's current attitude before it is pushed forward will cause damage to the chip during the chip tape and reel feeding process, issue the corresponding level of feeding control command, and execute the corresponding anti-damage feeding and pushing action.

2. The chip taping machine anti-damage feeding method based on computer vision according to claim 1, characterized in that: The specific steps in S1 include: S11. Deploy a fixed-focus industrial camera module and a three-point matrix light source on the feeding track plane of the chip taping machine. The specific deployment process is as follows: The working field of view and image projection area of ​​the industrial camera module are determined based on the feeding track plane of the chip taping machine. The optical axis of the industrial camera module is vertically aligned with the feeding track plane of the chip taping machine. At the same time, a three-point matrix light source is deployed directly above the feeding track surface of the chip taping machine. The three-point matrix light source includes three sets of LED arrays, which are located at the upper left, upper right and upper front, and illuminate the surface at an oblique angle. S12. Before the chip feeding and pushing operation of the chip tape and reel machine, the chips in the feeding track before they are pushed forward are photographed to form the original image data matrix of the chips before they are pushed forward. The specific process is as follows: During continuous illumination by the three-point matrix light source, the image acquisition controller of the industrial camera module, based on the feeding status signal of the chip tape machine, fixes the image acquisition time of the industrial camera module within the static time window of the chip before it is advanced. The industrial camera module then performs a complete exposure and readout operation within this static time window. After grayscale processing, the original image data matrix of the chip before it is advanced is formed. The original image data matrix Ghd... xy This indicates that the grayscale value of the corresponding pixel unit in the plane coordinates (x, y) is Ghd. i ; S13. Perform preprocessing operations on the original image data matrix of the chip before it is pushed forward. The preprocessing operations include mean filtering and contrast normalization.

3. The chip taping machine anti-damage feeding method based on computer vision according to claim 2, characterized in that: The specific steps in S2 include: S21. Based on the original image data matrix of the chip before advancement, determine the reflection interference coefficient of each pixel unit. The specific process is as follows: Feature recognition is performed on the original image data matrix of the chip before it is pushed forward in S12. The gray value of each pixel unit of the chip before it is pushed forward in the plane coordinate is extracted. Combined with the statistical mean algorithm, the average gray value of the pixels in the plane coordinate is obtained. The average grayscale value of each pixel in the planar coordinate system is correlated with the grayscale value of each pixel unit in the corresponding planar coordinate system. Based on the brightness response difference model under multi-angle light source illumination, and after dimensionless processing, the degree of reflective interference of each pixel unit in the planar coordinate system is analyzed to determine the reflective interference coefficient of each pixel unit. Specifically: In the formula, Xgr i This represents the reflection interference coefficient of the corresponding pixel unit, in Ghb. i This represents the grayscale value of the corresponding pixel unit. Represents the average grayscale value of pixels in a planar coordinate system. This represents the cosine value of the angle between the normal axis of the corresponding pixel unit and the direction of the incident light.

4. The chip taping machine anti-damage feeding method based on computer vision according to claim 3, characterized in that: S22. Based on the correlation between the original image data matrix of the chip before advancement and the reflection interference coefficient of each pixel unit in S21, and combined with the second-order gradient algorithm, analyze the overall image clarity and anti-reflection level of the chip before advancement under the current lighting conditions, so as to determine the image quality coefficient of the chip before advancement under the current lighting conditions, specifically: In the formula, Ghd represents the image quality coefficient of the chip before it is advanced under the current lighting conditions. xy Represents the original image data matrix. This represents the sum of the second-order grayscale derivatives of the corresponding pixel unit in the horizontal and vertical directions. Xgr represents the partial derivative sign. i This represents the reflection interference coefficient of the corresponding pixel unit, where n represents the number of pixel units.

5. A chip taping machine anti-damage feeding method based on computer vision according to claim 4, characterized in that: S23. Analyze whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and issue corresponding attitude recognition commands. The specific analysis process is as follows: If the image quality coefficient of the chip before advancement under the current lighting conditions exceeds the preset quality threshold, it indicates that the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and an attitude analysis signal is issued. If the image quality coefficient of the chip before advancement does not exceed the preset quality threshold under the current lighting conditions, it means that the image acquisition effect of the chip before advancement under the current lighting conditions is not suitable for subsequent chip attitude recognition. At this time, the intensity of the three-point matrix light source and the exposure time of the industrial camera module are adjusted, and the image acquisition operation of the chip before advancement is repeated.

6. A chip taping machine anti-damage feeding method based on computer vision according to claim 5, characterized in that: The specific steps of S3 include: S31. After receiving the attitude analysis signal, based on the original image data matrix of the chip before advancement, edge detection and closed contour extraction methods are used to identify the two-dimensional outer contour of the chip before advancement to determine the coordinates of the image center point. Based on the standard size parameters of the chip taping machine's feeding track plane, the coordinates of the track center point of the chip taping machine's feeding track plane are extracted. After calculating the Euclidean distance, the eccentricity distance of the chip before advancement is determined, specifically: In the formula, Dpx represents the eccentric distance of the chip before advancement, (x1-x2) represents the distance between the center of the track and the center of the image on the x-axis, (y1-y2) represents the distance between the center of the track and the center of the image on the y-axis, and R represents the scaling factor.

7. A chip taping machine anti-damage feeding method based on computer vision according to claim 6, characterized in that: S32. Perform pixel-level feature matching on the original image data matrix to extract the relative spatial tilt angle data matrix of the chip before advancement. The specific extraction process is as follows: By acquiring the left and right views from the original image data matrix separately and performing pixel-level feature matching between the view images, the position offset values ​​of the corresponding pixel units are extracted to construct a three-dimensional depth map of the chip before it is advanced. Extract the maximum gradient line segment from the high gradient region of the 3D depth map, and extract the angle between the maximum gradient line segment and the reference horizontal plane of the chip tape machine's feeding track. Obtain the relative spatial tilt angle data matrix of the chip before advancement, specifically in the form: Tα=[Tα x ,Tα y ,Tα z ]; In the formula, Tα x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y Tα represents the tilt angle of the chip in the y-axis direction before it is advanced. z This indicates the tilt angle of the chip in the z-axis direction before it is advanced.

8. A chip taping machine anti-damage feeding method based on computer vision according to claim 7, characterized in that: S33. Based on the correlation between the relative spatial tilt angle data matrix of the chip before advancement and the eccentricity distance, and combined with the image quality coefficient of the chip before advancement under the current lighting conditions, after dimensionless processing, the attitude deviation degree of the chip before advancement under the current lighting conditions is analyzed to determine the attitude deviation index of the chip before advancement under the current lighting conditions, specifically: In the formula, Zzt represents the feeding posture offset index of the chip tape-and-reel machine under the current illumination conditions, and Tα x Tα represents the tilt angle of the chip in the x-axis direction before it is advanced. y represents the tilt angle of the chip in the y-axis direction before advancement, Dpx represents the eccentricity distance of the chip before advancement, Xtz represents the image quality coefficient of the chip before advancement, and arctan(*) represents the tangent function.

9. A chip taping machine anti-damage feeding method based on computer vision according to claim 8, characterized in that: The specific steps of S4 include: S41. Based on the obtained attitude deviation index of the chip before advancement under the current lighting conditions, compare it with a preset deviation threshold to determine whether the chip's attitude before advancement under the current lighting conditions will cause damage to the chip during the chip tape and reel feeding process. Issue a corresponding level of feeding control command and execute the corresponding anti-damage feeding and advancing action, specifically: If the attitude deviation index of the chip before advancement under the current lighting conditions is greater than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a first-level feeding control command is issued. The chip tape and reel machine stops feeding and advancing operation and prompts the operator to reposition the chip. If the attitude deviation index of the chip before advancement under the current lighting conditions is equal to the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a secondary feeding control command is issued, and the chip tape and reel machine performs a flexible feeding and advancing operation. If the attitude deviation index of the chip before advancement under the current lighting conditions is less than the deviation threshold, it means that the attitude of the chip before advancement under the current lighting conditions will not cause damage to the chip during the feeding and advancing process of the chip tape and reel machine. At this time, a level 3 feeding control command is issued, and the chip tape and reel machine performs normal feeding and advancing operation.

10. A computer vision-based chip taping machine anti-damage feeding system, used to implement the computer vision-based chip taping machine anti-damage feeding method according to any one of claims 1 to 9, characterized in that: It includes a vision acquisition module, an image quality analysis module, a posture analysis module, and a material feeding control module; The vision acquisition module is based on a fixed-focus industrial camera module and a three-point matrix light source. Before the chip feeding and pushing operation of the chip tape machine, it takes pictures of the chip in the feeding track before it is pushed, forming the original image data matrix of the chip before it is pushed. The image quality analysis module, based on the original image data matrix and combined with the second-order gradient algorithm, determines the image quality coefficient of the chip before advancement under the current lighting conditions. This is used to determine whether the image acquisition effect of the chip before advancement under the current lighting conditions is suitable for subsequent chip attitude recognition, and to issue corresponding attitude recognition commands. The attitude analysis module is used to determine the eccentricity distance and relative spatial tilt angle data matrix of the chip before it is pushed forward, based on the original image data matrix after receiving the attitude analysis signal. Combined with the image quality coefficient, it analyzes the degree of attitude deviation of the chip before it is pushed forward, so as to determine the attitude deviation index of the chip before it is pushed forward. The feeding control module is used to determine whether the current attitude of the chip before it is pushed will cause damage to the chip during the feeding process of the chip tape machine, based on the attitude deviation index of the chip before it is pushed. It then issues a corresponding level of feeding control command and executes the corresponding anti-damage feeding action.

Images