A system and method for detecting appearance defects based on a robotic arm

Abstract

The present invention discloses a system and method for appearance defect detection based on a robot arm, and relates to the technical field of appearance defect detection. The detection system includes a control module, an analysis path module, a position correction module, and an instruction switching module. The present invention predicts the probability of appearance defects at various angles of the product and dynamically adjusts the detection path of the robot arm, so as to achieve the purpose of detecting appearance defects as soon as possible. When the appearance defect is detected, the detection operation of the robot arm can be stopped, reducing resource waste and improving detection efficiency; and the position deviation of the next product can be calculated to achieve the purpose of correcting the detection path of the robot arm, and at the same time control the node of instruction switching, and extend the use state of the robot arm after switching to a certain instruction as much as possible, reduce the operation of replacing the end effector of the robot arm, can save operation time, and can also reduce parts wear and increase the service life of parts.

Description

Technical Field

[0001] The present invention relates to the technical field of appearance defect detection, and in particular to an appearance defect detection system and method based on a robotic arm. Background Art

[0002] A robotic arm is a complex system with high precision, multiple inputs and multiple outputs, high nonlinearity, and strong coupling. Due to its unique operational flexibility, it has been widely used in industrial assembly, safety and explosion protection, and other fields. The robotic arm is a complex system with uncertainties such as parameter perturbations, external interference, and unmodeled dynamics. Therefore, the modeling model of the robotic arm also has uncertainties. For different tasks, it is necessary to plan the motion trajectory of the robotic arm joint space, so as to cascade and form the terminal posture.

[0003] When the robotic arm inspects the appearance of a product, a set of fixed instructions will be set to control the robotic arm to collect appearance images of the product at each angle in turn according to the instructions. However, for unqualified products, if the appearance defects of the product at that angle can be detected as soon as possible, there is no need to complete the full set of instructions, which reduces the waste of inspection resources and improves inspection efficiency. Summary of the invention

[0004] The purpose of the present invention is to provide a system and method for detecting appearance defects based on a robot arm to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for detecting appearance defects based on a robotic arm, the detection method comprising the following specific steps:

[0006] Step S1, issuing instructions to the robot arm to control the robot arm to collect appearance images of products, marking products with appearance defects in the appearance images as unqualified products, and removing them by using the robot arm;

[0007] Step S2, marking a record of an unqualified product as an unqualified record, extracting all unqualified records into an unqualified database, obtaining unqualified data of each unqualified record, and calculating the detection path of the robot arm when detecting the next product to be detected according to the unqualified data;

[0008] Step S3, collecting the position information of the next product to be detected, calculating the position deviation of the next product to be detected, and correcting the detection path of the next product to be detected according to the position deviation;

[0009] Step S4, collecting the detection data of the robot arm during the detection process, calculating the time node for issuing a switching instruction to the robot arm according to the detection data, and outputting the calculation result to the control system.

[0010] Furthermore, the step S1 includes:

[0011] The end effector of the robot arm includes an image acquisition device and a clamping device;

[0012] Sending an image acquisition instruction to the robot arm, the end effector controls the image acquisition device to acquire the appearance image of the product; when an appearance defect is detected in the acquired appearance image, the current product is marked as a defective product;

[0013] A rejection instruction is issued to the robot arm, and the end effector controls the clamping device to clamp the unqualified products and remove the unqualified products from the production line.

[0014] Furthermore, the step S2 includes:

[0015] Step S201, obtaining unqualified data of unqualified products, wherein the unqualified data includes appearance images of products, dividing the appearance images according to the angles at which the products are photographed, taking images photographed at the same angle as a node, arranging all nodes in the order of the movement path according to the movement path of the robot arm, extracting the angles of the products corresponding to the arranged nodes, and forming a movement path = {angle 1 → angle 2 → angle 3 ... → angle N}, wherein angle 1 → angle 2 → angle 3 ... → angle N means that the robot arm photographs the appearance image of the product from angle 1, then moves to angle 2 to photograph the appearance image of the product, moves to angle 3 to photograph the appearance image of the product ... moves to angle N to photograph the appearance image of the product;

[0016] Step S202: Mark the appearance image with appearance defects as the target image, obtain the angle of the product corresponding to the target image, and mark the angle as the target node, and obtain the probability F of each target node appearing. i , the calculation formula is F i =m i / M, where F i represents the probability of the i-th target node appearing, m i represents the number of occurrences of the i-th target node, and M represents the total number of all nodes;

[0017] Step S203: Obtain the probability F of each target node appearing i , according to the probability of the target node, the target node corresponding to the maximum probability is output, the angle corresponding to the output target node is extracted, and the angle is used as the starting point of the moving path to shoot the product appearance image at angle 1, and the probability corresponding to the adjacent angles of angle 1 is obtained. The angle with the largest corresponding probability among the adjacent angles is used as angle 2, and so on. All angles of the product are arranged to form a moving path, and the moving path is output as the detection path of the robot arm;

[0018] Step S204: Get the test result of the current product. If the current product is an unqualified product, set a time period T, get the unqualified data of all unqualified products in the time period T, extract the target nodes of all unqualified data, and calculate the frequency F of each target node. j =m j / T, where F j represents the frequency of the j-th target node appearing in time period T, m j represents the total number of times the j-th target node appears in the time period T; the target node corresponding to the highest frequency is output as the candidate angle, and the candidate angle is compared with the angle 1 output in step S203 to determine whether they are the same angle. If they are the same angle, the angle 1 is continued as the angle for taking the product appearance image at the beginning of the moving path of the next product; if they are not the same angle, the screening value W of the candidate angle and the angle 1 is calculated, and the calculation formula is W=tv*(F i *β1)*(F j *β2), where β1 and β2 are weight values, tv represents the proportion of the duration corresponding to time period T in the historical duration of detection; if the screening value of the alternative angle is greater than the screening value of angle one, the alternative angle will be used as the new angle one for shooting the product appearance image at the beginning of the moving path of the next product detection; if the screening value of the alternative angle is less than the screening value of angle one, angle one will continue to be used as the angle for shooting the product appearance image at the beginning of the moving path of the next product.

[0019] Furthermore, the step S3 includes:

[0020] Collect the top view of the current product and the next product to be inspected, take the corner points of the product as feature points, establish a coordinate system with the edge of the conveyor belt as the axis, and mark the coordinate point G of each feature point in the current product k (x, y), and the coordinate point of each feature point in the next product to be detected is H k (x, y), calculate the angle between each feature point and the edge of the conveyor belt, make the feature points in the next product to be detected correspond to the feature points in the current product one by one, and compare whether the angles between the two corresponding feature points and the edge of the conveyor belt are the same. If they are the same, the judgment is passed. If they are not the same, the feature point is used as the target feature point, and the angle of the product corresponding to the target feature point is output. The angle L of the current product and the angle P of the next product are obtained respectively. The difference between the angle P and the angle L is used as the position deviation degree, and the detection path of the next product to be detected is corrected according to the position deviation degree.

[0021] Further, the step S4 includes:

[0022] Obtain the test results of each product during the test process, extract all the nodes that a single qualified product passes through in the test path, and use the nodes that a single qualified product passes through when collecting appearance images as a complete node chain;

[0023] When an unqualified product appears, the target image corresponding to the appearance defect in the unqualified product is obtained, the target node corresponding to the target image is obtained, and the target nodes passed by the current unqualified product inspection are formed into an unqualified node chain. If the number of nodes in the unqualified node chain is less than the number of nodes in the complete node chain, no rejection instruction is output, but an image acquisition instruction is output to control the robot arm to continue to acquire the appearance image of the next product until a qualified product is acquired or until the distance between the acquired product and the current product is closest to the maximum movable distance range of the robot arm. Then, a rejection instruction is output and the robot arm is controlled to return to reject the current unqualified product. If the number of nodes in the unqualified node chain is equal to the number of nodes in the complete node chain, a rejection instruction is output to remove the current product from the production line.

[0024] A mechanical arm-based appearance defect detection system is applied to a mechanical arm-based appearance defect detection method, the detection system comprising a control module, an analysis path module, a position correction module and an instruction switching module;

[0025] The control module is used to issue instructions to the robotic arm to control the robotic arm to collect appearance images of products, mark products with appearance defects in the appearance images as unqualified products, and remove them using the robotic arm;

[0026] The analysis path module is used to obtain the unqualified data of each unqualified record, and calculate the detection path when the robot arm detects the next product to be detected according to the unqualified data;

[0027] The position correction module is used to calculate the position deviation of the next product to be detected, and to correct the detection path of the next product to be detected according to the position deviation;

[0028] The command switching module is used to calculate the time node for issuing command switching to the robot arm according to the detection data, and output the calculation result to the control system.

[0029] Furthermore, the control module includes an image acquisition device and a clamping device; the image acquisition device is used to acquire an appearance image of the product and determine whether the product is qualified; the clamping device is used to remove unqualified products from the production line.

[0030] Furthermore, the analysis path module includes an angle determination unit, an angle replacement unit and a path generation unit; the angle determination unit is used to use the angle of the product with the highest probability of appearance defects as the angle for starting to capture product appearance images in the detection path according to the angles of the divided products; the angle replacement unit is used to detect the angles of products corresponding to appearance defects in unqualified products in real time, and replace the angle for starting to capture product appearance images according to the frequency of appearance defects at the angle of the product in the current time period; the path generation unit is used to arrange the angles of all products according to the probability of appearance defects in the angles of the remaining products based on the determined angle for starting to capture product appearance images, so as to form a detection path.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The present invention predicts the probability of appearance defects at various angles of the product and dynamically adjusts the detection path of the robot arm, thereby achieving the purpose of detecting appearance defects as soon as possible. When appearance defects are detected, the detection operation of the robot arm can be stopped, reducing resource waste and improving detection efficiency; and the position deviation of the next product can be calculated to achieve the purpose of correcting the detection path of the robot arm. At the same time, the command switching node is controlled to extend the use state of the robot arm after switching to a certain command as much as possible, reducing the operation of replacing the end effector of the robot arm, which can not only save operation time, but also reduce parts wear and increase the service life of parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The present invention is a schematic flow chart of a method for detecting appearance defects based on a robotic arm. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Embodiment 1: Figure 1 As shown, the present invention provides an appearance defect detection method based on a robotic arm, and the detection method comprises the following specific steps:

[0036] Step S1, issuing instructions to the robot arm to control the robot arm to collect appearance images of products, marking products with appearance defects in the appearance images as unqualified products, and removing them by using the robot arm;

[0037] Specifically, the step S1 includes:

[0038] The end effector of the robot arm includes an image acquisition device and a clamping device;

[0039] Sending an image acquisition instruction to the robot arm, the end effector controls the image acquisition device to acquire the appearance image of the product; when an appearance defect is detected in the acquired appearance image, the current product is marked as a defective product;

[0040] A rejection instruction is issued to the robot arm, and the end effector controls the clamping device to clamp the unqualified products and remove the unqualified products from the production line.

[0041] Step S2, marking a record of an unqualified product as an unqualified record, extracting all unqualified records into an unqualified database, obtaining unqualified data of each unqualified record, and calculating the detection path of the robot arm when detecting the next product to be detected according to the unqualified data;

[0042] Specifically, the step S2 includes: step S201, obtaining unqualified data of unqualified products, the unqualified data including appearance images of products, dividing the appearance images according to the angles at which the products are photographed, taking images photographed at the same angle as a node, arranging all nodes in the order of the moving path according to the moving path of the robot arm, extracting the angles of the products corresponding to the arranged nodes, and forming a moving path = {angle one → angle two → angle three ... → angle N}, wherein angle one → angle two → angle three ... → angle N means that the robot arm photographs the appearance image of the product from angle one, then moves to angle two to photograph the appearance image of the product, moves to angle three to photograph the appearance image of the product ... and moves to angle N to photograph the appearance image of the product;

[0043] Step S202: Mark the appearance image with appearance defects as the target image, obtain the angle of the product corresponding to the target image, and mark the angle as the target node, and obtain the probability F of each target node appearing. i , the calculation formula is F i =m i / M, where F i represents the probability of the i-th target node appearing, m i represents the number of occurrences of the i-th target node, and M represents the total number of all nodes;

[0044] Step S203: Obtain the probability F of each target node appearing i, according to the probability of the target node, the target node corresponding to the maximum probability is output, the angle corresponding to the output target node is extracted, and the angle is used as the starting point of the moving path to shoot the product appearance image at angle 1, and the probability corresponding to the adjacent angles of angle 1 is obtained. The angle with the largest corresponding probability among the adjacent angles is used as angle 2, and so on. All angles of the product are arranged to form a moving path, and the moving path is output as the detection path of the robot arm;

[0045] Step S204: Get the test result of the current product. If the current product is an unqualified product, set a time period T, get the unqualified data of all unqualified products in the time period T, extract the target nodes of all unqualified data, and calculate the frequency F of each target node. j =m j / T, where F j represents the frequency of the j-th target node appearing in time period T, m j represents the total number of times the j-th target node appears in the time period T; the target node corresponding to the highest frequency is output as the candidate angle, and the candidate angle is compared with the angle 1 output in step S203 to determine whether they are the same angle. If they are the same angle, the angle 1 is continued as the angle for taking the product appearance image at the beginning of the moving path of the next product; if they are not the same angle, the screening value W of the candidate angle and the angle 1 is calculated, and the calculation formula is W=tv*(F i *β1)*(F j *β2), where β1 and β2 are weight values, tv represents the proportion of the duration corresponding to time period T in the historical duration of detection; if the screening value of the alternative angle is greater than the screening value of angle one, the alternative angle will be used as the new angle one for shooting the product appearance image at the beginning of the moving path of the next product detection; if the screening value of the alternative angle is less than the screening value of angle one, angle one will continue to be used as the angle for shooting the product appearance image at the beginning of the moving path of the next product.

[0046] For example, it is necessary to capture appearance images from the left, right, front, back and top of a product. The probability of appearance defects appearing in the appearance images corresponding to the left, right, front, back and top can be calculated, and the surface with a high probability of appearance defects appearing can be used as the starting angle for collecting appearance images. This can increase the probability of collecting appearance defects and achieve the purpose of collecting appearance defects as soon as possible.

[0047] Step S3, collecting the position information of the next product to be detected, calculating the position deviation of the next product to be detected, and correcting the detection path of the next product to be detected according to the position deviation;

[0048] Specifically, step S3 includes:

[0049] Collect the top view of the current product and the next product to be inspected, take the corner points of the product as feature points, establish a coordinate system with the edge of the conveyor belt as the axis, and mark the coordinate point G of each feature point in the current product k (x, y), and the coordinate point of each feature point in the next product to be detected is H k (x, y), calculate the angle between each feature point and the edge of the conveyor belt, make the feature points in the next product to be detected correspond to the feature points in the current product one by one, and compare whether the angles between the two corresponding feature points and the edge of the conveyor belt are the same. If they are the same, the judgment is passed. If they are not the same, the feature point is used as the target feature point, and the angle of the product corresponding to the target feature point is output. The angle L of the current product and the angle P of the next product are obtained respectively. The difference between the angle P and the angle L is used as the position deviation degree, and the detection path of the next product to be detected is corrected according to the position deviation degree.

[0050] Specifically, for example, if the next product is offset by 30 degrees, then in the originally set control instruction, the first collected angle is increased by an offset of 30 degrees.

[0051] Step S4: collect the detection data of the robot arm during the detection process, calculate the time node for issuing a command switch to the robot arm according to the detection data, and output the calculation result to the control system.

[0052] Specifically, step S4 includes:

[0053] Obtain the test results of each product during the test process, extract all the nodes that a single qualified product passes through in the test path, and use the nodes that a single qualified product passes through when collecting appearance images as a complete node chain;

[0054] When an unqualified product appears, the target image corresponding to the appearance defect in the unqualified product is obtained, the target node corresponding to the target image is obtained, and the target nodes passed by the current unqualified product inspection are formed into an unqualified node chain. If the number of nodes in the unqualified node chain is less than the number of nodes in the complete node chain, no rejection instruction is output, but an image acquisition instruction is output to control the robot arm to continue to acquire the appearance image of the next product until a qualified product is acquired or until the distance between the acquired product and the current product is closest to the maximum movable distance range of the robot arm. Then, a rejection instruction is output and the robot arm is controlled to return to reject the current unqualified product. If the number of nodes in the unqualified node chain is equal to the number of nodes in the complete node chain, a rejection instruction is output to remove the current product from the production line.

[0055] For example, if a certain surface has appearance defects continuously within the current period of time, the angle of the surface will be used as the starting angle for detection and replaced. Because according to the probability calculation, the surface with appearance defects in the previous batch of products may be different from the surface with appearance defects in this batch of products, but according to historical records, the surface corresponding to the appearance defects of the previous batch of products is still collected, which will lead to large errors. Therefore, it is necessary to update and replace according to the surface corresponding to the recent appearance defects.

[0056] Embodiment 2: A mechanical arm-based appearance defect detection system, which applies a mechanical arm-based appearance defect detection method, the detection system includes a control module, an analysis path module, a position correction module and an instruction switching module;

[0057] The control module is used to issue instructions to the robotic arm, control the robotic arm to collect appearance images of products, mark products with appearance defects in the appearance images as unqualified products, and use the robotic arm to remove them.

[0058] Specifically, the control module includes an image acquisition device and a clamping device; the image acquisition device is used to acquire the appearance image of the product and determine whether the product is qualified; the clamping device is used to remove unqualified products from the production line.

[0059] The analysis path module is used to obtain the unqualified data of each unqualified record, and calculate the detection path when the robot arm detects the next product to be detected according to the unqualified data.

[0060] Specifically, the analysis path module includes an angle determination unit, an angle replacement unit and a path generation unit; the angle determination unit is used to use the angle of the product with the highest probability of appearance defects as the angle for starting to capture product appearance images in the detection path according to the angles of the divided products; the angle replacement unit is used to detect the angles of products corresponding to appearance defects in unqualified products in real time, and replace the angle for starting to capture product appearance images according to the frequency of appearance defects at the angle of the product in the current time period; the path generation unit is used to arrange the angles of all products according to the probability of appearance defects in the angles of the remaining products according to the determined angle for starting to capture product appearance images, so as to form a detection path.

[0061] The position correction module is used to calculate the position deviation of the next product to be detected, and to correct the detection path of the next product to be detected according to the position deviation;

[0062] The command switching module is used to calculate the time node for issuing command switching to the robot arm according to the detection data, and output the calculation result to the control system.

[0063] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for detecting appearance defects based on a robotic arm, characterized in that: The detection method comprises the following specific steps: Step S1, issuing instructions to the robot arm to control the robot arm to collect appearance images of products, marking products with appearance defects in the appearance images as unqualified products, and removing them by using the robot arm; Step S2, marking a record of an unqualified product as an unqualified record, extracting all unqualified records into an unqualified database, obtaining unqualified data of each unqualified record, and calculating the detection path of the robot arm when detecting the next product to be detected according to the unqualified data; Step S3, collecting the position information of the next product to be detected, calculating the position deviation of the next product to be detected, and correcting the detection path of the next product to be detected according to the position deviation; Step S4, collecting detection data of the robot arm during the detection process, calculating the time node for issuing a command switch to the robot arm according to the detection data, and outputting the calculation result to the control system; The step S2 comprises: Step S201, obtaining unqualified data of unqualified products, wherein the unqualified data includes appearance images of products, dividing the appearance images according to the angles at which the products are photographed, taking images photographed at the same angle as a node, arranging all nodes in the order of the movement path according to the movement path of the robot arm, extracting the angles of the products corresponding to the arranged nodes, and forming a movement path = {angle 1 → angle 2 → angle 3 ... → angle N}, wherein angle 1 → angle 2 → angle 3 ... → angle N means that the robot arm photographs the appearance image of the product from angle 1, then moves to angle 2 to photograph the appearance image of the product, moves to angle 3 to photograph the appearance image of the product ... moves to angle N to photograph the appearance image of the product; Step S202: Mark the appearance image with appearance defects as the target image, obtain the angle of the product corresponding to the target image, and mark the angle as the target node, and obtain the probability F of each target node appearing. i , the calculation formula is F i =m i / M, where F i represents the probability of the i-th target node appearing, m i represents the number of occurrences of the i-th target node, and M represents the total number of all nodes; Step S203: Obtain the probability F of each target node appearing i , according to the probability of the target node, the target node corresponding to the maximum probability is output, the angle corresponding to the output target node is extracted, and the angle is used as the starting point of the moving path to shoot the product appearance image at angle 1, and the probability corresponding to the adjacent angles of angle 1 is obtained. The angle with the largest corresponding probability among the adjacent angles is used as angle 2, and so on. All angles of the product are arranged to form a moving path, and the moving path is output as the detection path of the robot arm; Step S204: Get the test result of the current product. If the current product is an unqualified product, set a time period T, get the unqualified data of all unqualified products in the time period T, extract the target nodes of all unqualified data, and calculate the frequency F of each target node. j =m j / T, where F j represents the frequency of the j-th target node appearing in time period T, m j represents the total number of times the j-th target node appears in the time period T; the target node corresponding to the highest frequency is output as the candidate angle, and the candidate angle is compared with the angle 1 output in step S203 to determine whether they are the same angle. If they are the same angle, the angle 1 is continued as the angle for taking the product appearance image at the beginning of the moving path of the next product; if they are not the same angle, the screening value W of the candidate angle and the angle 1 is calculated, and the calculation formula is W=tv*(F i *β1)*(F j *β2), where β1 and β2 are weight values, and tv represents the proportion of the duration corresponding to time period T in the historical duration of the detection; if the screening value of the alternative angle is greater than the screening value of angle 1, the alternative angle is used as the new angle 1 for shooting the product appearance image at the beginning of the moving path of the next product detection; if the screening value of the alternative angle is less than the screening value of angle 1, angle 1 continues to be used as the angle for shooting the product appearance image at the beginning of the moving path of the next product; The step S3 comprises: Collect the top view of the current product and the next product to be inspected, take the corner points of the product as feature points, establish a coordinate system with the edge of the conveyor belt as the axis, and mark the coordinate point G of each feature point in the current product k (x, y), and the coordinate point of each feature point in the next product to be detected is H k (x, y), calculate the angle between each feature point and the edge of the conveyor belt, match the feature points in the next product to be detected with the feature points in the current product one by one, and compare whether the angles between the two corresponding feature points and the edge of the conveyor belt are the same. If they are the same, the judgment is passed. If they are not the same, the feature point is used as the target feature point, and the angle of the product corresponding to the target feature point is output. The angle L of the current product and the angle P of the next product are obtained respectively. The difference between the angle P and the angle L is used as the position deviation degree, and the detection path of the next product to be detected is corrected according to the position deviation degree; The step S4 comprises: Obtain the test results of each product during the test process, extract all the nodes that a single qualified product passes through in the test path, and use the nodes that a single qualified product passes through when collecting appearance images as a complete node chain; When an unqualified product appears, the target image corresponding to the appearance defect in the unqualified product is obtained, the target node corresponding to the target image is obtained, and the target nodes passed by the current unqualified product inspection are formed into an unqualified node chain. If the number of nodes in the unqualified node chain is less than the number of nodes in the complete node chain, no rejection instruction is output, but an image acquisition instruction is output to control the robot arm to continue to acquire the appearance image of the next product until a qualified product is acquired or until the distance between the acquired product and the current product is closest to the maximum movable distance range of the robot arm. Then, a rejection instruction is output and the robot arm is controlled to return to reject the current unqualified product. If the number of nodes in the unqualified node chain is equal to the number of nodes in the complete node chain, a rejection instruction is output to remove the current product from the production line.

2. The appearance defect detection method based on a robotic arm according to claim 1, characterized in that: The step S1 comprises: The end effector of the robot arm includes an image acquisition device and a clamping device; Sending an image acquisition instruction to the robot arm, the end effector controls the image acquisition device to acquire the appearance image of the product; when an appearance defect is detected in the acquired appearance image, the current product is marked as a defective product; A rejection instruction is issued to the robot arm, and the end effector controls the clamping device to clamp the unqualified products and remove the unqualified products from the production line.

3. A robot-based appearance defect detection system, which is applied to a robot-based appearance defect detection method according to any one of claims 1 to 2, characterized in that: The detection system includes a control module, an analysis path module, a position correction module and an instruction switching module; The control module is used to issue instructions to the robotic arm to control the robotic arm to collect appearance images of products, mark products with appearance defects in the appearance images as unqualified products, and remove them using the robotic arm; The analysis path module is used to obtain the unqualified data of each unqualified record, and calculate the detection path when the robot arm detects the next product to be detected according to the unqualified data; The position correction module is used to calculate the position deviation of the next product to be detected, and to correct the detection path of the next product to be detected according to the position deviation; The command switching module is used to calculate the time node for issuing command switching to the robot arm according to the detection data, and output the calculation result to the control system.

4. The appearance defect detection system based on a robotic arm according to claim 3, characterized in that: The control module includes an image acquisition device and a clamping device; the image acquisition device is used to acquire the appearance image of the product and determine whether the product is qualified; the clamping device is used to remove unqualified products from the production line.

5. The appearance defect detection system based on a robotic arm according to claim 3, characterized in that: The analysis path module includes an angle determination unit, an angle replacement unit and a path generation unit; the angle determination unit is used to use the angle of the product with the highest probability of appearance defects as the angle for starting to capture product appearance images in the detection path according to the angles of the divided products; the angle replacement unit is used to detect the angles of products corresponding to appearance defects in unqualified products in real time, and replace the angle for starting to capture product appearance images according to the frequency of appearance defects at the angle of the product in the current time period; the path generation unit is used to arrange the angles of all products according to the probability of appearance defects in the angles of the remaining products according to the determined angle for starting to capture product appearance images, so as to form a detection path.

Images