Connector quality inspection system based on multi-angle vision

Abstract

The invention discloses a connector quality inspection system based on multi-angle vision, and relates to the technical field of connectors, and the technical scheme is characterized by comprising the following steps: carrying out region division on a multi-angle vision image set of a connector at different illumination angles to obtain a key detection region block and a peripheral auxiliary region block; processing and analyzing the key detection area blocks and the peripheral auxiliary area blocks to obtain quality abnormity early warning information or continuous inspection information; collecting appearance characteristic dynamic change data of the key detection area block and the peripheral auxiliary area block according to the continuous inspection information, and judging a first appearance state of the key detection area block in combination with a connector standard appearance parameter library; correcting the first appearance state of the key detection area block based on a feature association relationship between the key detection area block and the peripheral auxiliary area block to obtain a second appearance state; the effect is that a powerful basis is provided for quality control of the connector.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of connectors, more particularly, it relates to a connector quality inspection system based on multi-angle vision. BACKGROUND

[0002] In the field of connector quality inspection, the traditional inspection method only obtains images from one or a few fixed illumination angles, and scratches, local color differences and the like at specific angles are easily missed, resulting in insufficient inspection accuracy.

[0003] At the same time, the traditional inspection method basically only focuses on the appearance and size of the connector and the like, and easily ignores the influence of environmental factors in the actual assembly scene of the connector on the quality of the connector. However, in fact, environmental factors have an effect on the performance and service life of the connector during long-term use, and thus cause quality failures.

[0004] Therefore, the lack of consideration of environmental factors makes the traditional inspection result unable to fully reflect the real quality status of the connector, which is not conducive to effective control of the quality of the connector. SUMMARY

[0005] In view of the deficiencies in the prior art, the purpose of the present application is to provide a connector quality inspection system based on multi-angle vision.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme: A connector quality inspection system based on multi-angle vision, comprising: A division module: performing regional division on a multi-angle vision image set of the connector at different illumination angles to obtain key detection area blocks and peripheral auxiliary area blocks; An analysis module: processing and analyzing the key detection area blocks and the peripheral auxiliary area blocks to obtain quality abnormality early warning information or continue inspection information; A judgment module: collecting appearance feature dynamic change data of the key detection area blocks and the peripheral auxiliary area blocks according to the continue inspection information, and combining a connector standard appearance parameter library to judge a first appearance state of the key detection area blocks; A first processing module: correcting the first appearance state of the key detection area blocks based on the feature correlation relationship of the key detection area blocks and the peripheral auxiliary area blocks to obtain a second appearance state; and processing the second appearance state to obtain a first quality risk value of the key detection area blocks; A second processing module: constructing an environmental failure correlation model according to environmental interference parameters and historical quality failure data of a corresponding assembly scene of the connector, and obtaining a second quality risk value of the key detection area blocks by the environmental failure correlation model from the appearance feature dynamic change data and the first quality risk value; The output module: according to the first quality risk value and the second quality risk value, the connector is comprehensively quality inspected to obtain a comprehensive quality inspection result.

[0007] Preferably, the key detection area block and the surrounding auxiliary area block are processed and analyzed to obtain quality abnormal early warning information or continue to inspect information, specifically including the following steps: The appearance feature information of the key detection area block and the surrounding auxiliary area block and the feature correlation coefficient therebetween are extracted; The feature correlation reference interval value is obtained by processing and analyzing the image data of the connector under historical conditions; The current feature correlation coefficient set of the key detection area block and the surrounding auxiliary area block is counted in the current detection period; The current feature correlation coefficient set is compared with the feature correlation reference interval value to obtain quality abnormal early warning information or continue to inspect information.

[0008] Preferably, the appearance feature information of the key detection area block and the surrounding auxiliary area block and the feature correlation coefficient therebetween are extracted, specifically including the following steps: The appearance feature information of the key detection area block and the surrounding auxiliary area block is extracted, wherein the appearance feature information includes color parameters, contour size deviation values and surface defect area ratios; The feature correlation coefficient set of the key detection area block and the surrounding auxiliary area block is calculated, wherein the feature correlation coefficient set includes color similarity coefficients, size correlation coefficients and defect influence coefficients.

[0009] Preferably, the feature correlation reference interval value is obtained by processing and analyzing the image data of the connector under historical conditions, specifically including the following steps: The multi-angle visual image data of the connector under historical normal working conditions is extracted to obtain a historical normal feature correlation coefficient set; The multi-angle visual image data of the connector under historical quality failure conditions is extracted to obtain a historical failure feature correlation coefficient set; The feature correlation reference interval value is determined according to the distribution range of the historical normal feature correlation coefficient set and the abnormal threshold of the historical failure feature correlation coefficient set.

[0010] Preferably, the current feature correlation coefficient set is compared with the feature correlation reference interval value to obtain quality abnormal early warning information or continue to inspect information, specifically including the following steps: The current feature correlation coefficient set is compared with the corresponding feature correlation reference interval value; If the current feature correlation coefficient set exceeds the feature correlation reference interval value, it is determined that the area block corresponding to the correlation reference interval value is exceeded, and quality abnormal early warning information is generated; If the current feature correlation coefficient set is within the feature correlation reference interval value, a difference value data set is calculated by calculating the difference between the current feature correlation coefficient set and the upper limit of the corresponding feature correlation reference interval value; If the difference value data set is greater than or equal to the preset difference value threshold, continue to generate verification information.

[0011] Preferably, according to the continue to verify the information acquisition key detection area block and the surrounding auxiliary area block appearance feature dynamic change data, specifically including the following steps: According to the detection time length requirement of the continue to verify the information to set the preset detection time length; Interval acquisition key detection area block and the surrounding auxiliary area block appearance feature dynamic change data within the preset detection time length, wherein the appearance feature dynamic change data includes color parameter, contour size deviation value and surface defect area ratio data.

[0012] Preferably, the appearance feature dynamic change data is combined with the connector standard appearance parameter library to judge the first appearance state of the key detection area block, specifically including the following steps: Establish a connector standard appearance parameter library; wherein the connector standard appearance parameter library includes the standard color range, the standard size deviation threshold and the standard defect area ratio of the connector key detection area block under different working time length; Through the time series prediction model to judge the trend of the appearance feature dynamic change data, combined with the standard parameter of the corresponding time length in the connector standard appearance parameter library to judge the first appearance state of the key detection area block after the preset detection time length; wherein the first appearance state includes color determination value, size deviation determination value and defect area ratio determination value.

[0013] Preferably, based on the feature correlation relationship between the key detection area block and the surrounding auxiliary area block, the first appearance state of the key detection area block is corrected to obtain the second appearance state; the second appearance state is processed to obtain the first quality risk value of the key detection area block, specifically including the following steps: According to the historical data to determine the feature correlation weight between the key detection area block and the surrounding auxiliary area block, wherein the feature correlation weight includes color correlation weight, size correlation weight, defect correlation weight; Multiply the first appearance state of the surrounding auxiliary area block by the corresponding feature correlation weight to obtain the weight correction value; According to the weight correction value, the first appearance state of the key detection area block is corrected to obtain the second appearance state; Establish an appearance state quality risk mapping model; The determination value, size deviation determination value and defect area ratio determination value of the second appearance state are input into the appearance state quality risk mapping model to obtain the first quality risk value of the key detection area block.

[0014] Preferably, the environment interference parameters of the corresponding assembly scene of the connector and the historical quality failure data are used to construct an environment failure correlation model, the dynamic change data of the appearance feature is combined with the first quality risk value to obtain the second quality risk value of the key detection area block through the environment failure correlation model, and the specific steps include the following steps: The environment interference parameters of the connector in the corresponding assembly scene are collected, wherein the environment interference parameters include environmental temperature, humidity, dust concentration and vibration frequency; The historical quality failure data of the connector under the environment interference parameters are collected, and an environment failure database is established; The environment failure database is trained by using a machine learning algorithm to obtain an environment failure correlation model; The environment interference parameters of the current assembly scene are input into the environment failure correlation model to obtain an environment influence coefficient, and the second quality risk value of the key detection area block is obtained according to the environment influence coefficient and the first quality risk value.

[0015] Preferably, the comprehensive quality inspection result of the connector is obtained according to the first quality risk value and the second quality risk value, and the specific steps include the following steps: The first-level risk threshold and the second-level risk threshold are preset, wherein the first-level risk threshold is less than the second-level risk threshold; If the first quality risk value and the second quality risk value are both lower than the first-level risk threshold, it is determined that the quality of the connector is qualified; If any of the first quality risk value and the second quality risk value is between the first-level risk threshold and the second-level risk threshold, it is determined that the connector needs to be repaired and re-inspected; If any of the first quality risk value and the second quality risk value is higher than the second-level risk threshold, it is determined that the quality of the connector is unqualified.

[0016] Compared with the prior art, the present application has the following advantages: The division module divides the multi-angle visual image set of the connector under different illumination angles to obtain key detection area blocks and peripheral auxiliary area blocks, which can cover the core and peripheral areas of the connector, and ensure that important parts are not missed in subsequent analysis. The analysis module processes and analyzes these area blocks, so as to give quality abnormal early warning information or continue to test information in time, realize preliminary and rapid screening of the quality of the connector, and avoid the flow of connectors with quality problems into subsequent links. The judgment module collects dynamic change data of the appearance characteristics according to the continue to test information, and judges the first appearance state in combination with the connector standard appearance parameter library, so as to grasp the appearance standard of the connector at different stages by means of the standard parameter library, so that the appearance state judgment is more scientific and normative. The first processing module corrects the first appearance state based on the feature correlation relationship to obtain the second appearance state, so that the appearance state is more in line with the actual situation, and the calculation of the quality risk value is more accurate, and the quality risk of the connector in the appearance level can be quantified. The second processing module constructs an environmental failure correlation model to obtain a second quality risk value in combination with the environmental interference parameters of the assembly scene and the historical quality failure data. Finally, the output module performs comprehensive quality inspection according to the two quality risk values, comprehensively judges the quality condition of the connector from two dimensions of the connector itself and environmental influence, provides a strong basis for the quality control of the connector, and reduces the loss caused by quality problems. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A schematic diagram of a connector quality inspection system based on multi-angle vision is provided for the present application; Figure 2 A step diagram for calculating feature correlation reference interval values in the connector quality inspection system based on multi-angle vision is provided for the present application; Figure 3 A feature correlation coefficient calculation principle diagram is provided for the present application; Figure 4 A quality risk determination logic diagram is provided for the present application; Figure 5 A schematic diagram of connector multi-angle image acquisition and area division is provided for the present application; Figure 6 An integrated schematic diagram of quality risk comprehensive determination is provided for the present application. DETAILED DESCRIPTION

[0018] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0019] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be appreciated that the present application can be practiced in a variety of ways beyond the specific details set forth herein without departing from the scope and spirit of the present application. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application.

[0020] It is also noted that, as used herein, "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one implementation of the present application. The appearances of the phrase "in one embodiment" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.

[0021] Referring to Figures 1-4 as shown.

[0022] Embodiments further illustrate a connector quality inspection system based on multi-angle vision proposed by the present application.

[0023] A connector quality inspection system based on multi-angle vision, comprising: a division module: dividing a multi-angle vision image set of the connector at different illumination angles into key detection area blocks and peripheral auxiliary area blocks through regional division; The multi-angle vision image set is divided into key detection area blocks and peripheral auxiliary area blocks by judging the features of each region in the multi-angle vision image set, such as color distribution, texture structure, and geometric shape. The key detection area blocks are the core part of the connector quality detection, such as the contact part and the core conductive area of the connector, which directly relates to whether the main function and performance of the connector meet the requirements; the peripheral auxiliary area blocks are the shell edge and non-core connection area of the connector, which can provide auxiliary information for the quality judgment of the key detection area blocks.

[0024] an analysis module: processing and analyzing the key detection area blocks and the peripheral auxiliary area blocks to obtain quality abnormal warning information or continue inspection information; a judgment module: collecting dynamic change data of the appearance features of the key detection area blocks and the peripheral auxiliary area blocks according to the continue inspection information, and combining a connector standard appearance parameter library to judge a first appearance state of the key detection area blocks; a first processing module: correcting the first appearance state of the key detection area blocks based on the feature correlation relationship of the key detection area blocks and the peripheral auxiliary area blocks to obtain a second appearance state; and processing the second appearance state to obtain a first quality risk value of the key detection area blocks; The second processing module: according to the environmental interference parameters of the connector corresponding to the assembly scene and the historical quality failure data, an environmental failure correlation model is constructed, and the second quality risk value of the key detection area block is obtained through the environmental failure correlation model from the appearance feature dynamic change data and the first quality risk value; The output module: according to the first quality risk value and the second quality risk value, the comprehensive quality inspection result of the connector is obtained.

[0025] The quality abnormality early warning information or the continuous inspection information is obtained by processing and analyzing the key detection area block and the surrounding auxiliary area block, specifically including the following steps: The appearance feature information of the key detection area block and the surrounding auxiliary area block and the feature correlation coefficient therebetween are extracted; The feature correlation reference interval value is obtained by processing and analyzing the image data of the connector under the historical condition; The current feature correlation coefficient set of the key detection area block and the surrounding auxiliary area block is counted in the current detection period; The quality abnormality early warning information or the continuous inspection information is obtained by comparing and judging the current feature correlation coefficient set and the feature correlation reference interval value.

[0026] The appearance feature information of the key detection area block and the surrounding auxiliary area block and the feature correlation coefficient therebetween are extracted, specifically including the following steps: The appearance feature information of the key detection area block and the surrounding auxiliary area block is extracted, wherein the appearance feature information includes color parameters, contour size deviation values and surface defect area proportion; The feature correlation coefficient set of the key detection area block and the surrounding auxiliary area block is calculated, wherein the feature correlation coefficient set includes color similarity coefficients, size correlation coefficients and defect influence coefficients.

[0027] The appearance feature information is extracted to obtain color parameters, contour size deviation values and surface defect area proportions of the key detection area block and the peripheral auxiliary area block. For example, taking a connector as an example, the key detection area is the core metal contact part, and the peripheral auxiliary area is the plastic shell area around the metal contact part. With the help of image analysis technology, the color parameters of the metal contact part, that is, the color values corresponding to the specific metal luster, are obtained; the deviation value of the actual contour size of the metal contact part from the standard size is measured; and the proportion of the defect area in the area when there is a scratch or other defects on the surface of the metal contact part is calculated. At the same time, the color parameters of the plastic shell area, that is, the color values of the specific color tone of the plastic, are obtained; the deviation value of the actual contour size of the plastic shell from the standard shell size is obtained; and the proportion of the area of the defects such as small dents on the surface of the plastic shell in the shell area is obtained. The color similarity coefficient, the size correlation coefficient and the defect influence coefficient of the key detection area block and the peripheral auxiliary area block are calculated. Taking this connector as an example, the color similarity coefficient is used to measure the degree of similarity of the color of the metal contact part and the plastic shell; the size correlation coefficient can reflect the correlation between the size of the metal contact part and the size of the plastic shell, such as whether the size of the shell is suitable for the size of the contact part; and the defect influence coefficient reflects the influence degree of the defects of the plastic shell on the metal contact part, such as whether the dents on the shell will cause the contact part to be offset and thus affect its function.

[0028] The image data of the connector is processed and analyzed under historical conditions to obtain a feature correlation reference interval value, specifically including the following steps: Extracting multi-angle visual image data of the connector under historical normal working conditions and obtaining a historical normal feature correlation coefficient set; Extracting multi-angle visual image data of the connector under historical quality failure conditions and obtaining a historical failure feature correlation coefficient set; Determining the feature correlation reference interval value according to the distribution range of the historical normal feature correlation coefficient set and the abnormal threshold of the historical failure feature correlation coefficient set.

[0029] The application extracts multi-angle visual image data of the connector in the historical normal working state, and obtains a historical normal feature correlation coefficient set by processing the image data. For example, a large number of visual images of the connector in the normal working state under different illumination angles are collected, the color similarity coefficient, size correlation coefficient and defect influence coefficient of the key detection area block and the peripheral auxiliary area block in the images are calculated, and thus the historical normal feature correlation coefficient set is formed. The multi-angle visual image data of the connector in the historical quality failure state is extracted, and the images in the failure state are also processed to obtain a historical failure feature correlation coefficient set. For example, the visual images of the connector with poor contact, damaged shell and other quality failures under different illumination angles are collected, the related features of the key detection area block and the peripheral auxiliary area block are judged, and the corresponding feature correlation coefficient is calculated to form the historical failure feature correlation coefficient set. The feature correlation reference interval value is determined according to the distribution range of the historical normal feature correlation coefficient set and the abnormal threshold of the historical failure feature correlation coefficient set. For example, the color similarity coefficient in the historical normal feature correlation coefficient set is between 0.8 and 0.95, and presents a certain distribution range; and when the color similarity coefficient in the historical failure feature correlation coefficient set is lower than 0.7, it corresponds to the quality failure of the connector, and 0.7 is taken as the abnormal threshold. The reasonable feature correlation reference interval value is determined by comprehensively considering the distribution range of the historical normal feature correlation coefficient set and the abnormal threshold of the historical failure feature correlation coefficient set, and is used for subsequent judgment of whether the feature correlation coefficient of the current connector is normal.

[0030] The quality abnormality early warning information or the continue inspection information is obtained by comparing the current feature correlation coefficient set with the feature correlation reference interval value, and specifically includes the following steps: The current feature correlation coefficient set is compared with the corresponding feature correlation reference interval value; If the current feature correlation coefficient set exceeds the feature correlation reference interval value, the area block corresponding to the exceeding correlation reference interval value is determined, and the quality abnormality early warning information is generated; If the current feature correlation coefficient set is within the feature correlation reference interval value, the difference value data set is obtained by calculating the difference between the current feature correlation coefficient set and the upper limit of the corresponding feature correlation reference interval value; If the difference value data set is greater than or equal to the preset difference threshold, the continue inspection information is generated.

[0031] The application first compares the current feature correlation coefficient set with the corresponding feature correlation reference interval value. Assuming that the normal interval of the color similarity coefficient in the feature correlation reference interval value is set to 0.8 to 0.95, if the color similarity coefficient of the current key detection area block and the surrounding auxiliary area block is 0.75 at this time, this value is obviously out of the interval range of 0.8 to 0.95, it is determined that the area block corresponding to the correlation reference interval value, that is, the key detection area block or the surrounding auxiliary area block, exists abnormal situation, and then generates quality abnormality early warning information, so as to prompt that there may be quality problems.

[0032] If the current feature correlation coefficient set is within the feature correlation reference interval value, for example, the color similarity coefficient is 0.9, which is within the interval of 0.8 to 0.95, it is necessary to calculate the difference value between the current feature correlation coefficient set and the upper limit of the corresponding feature correlation reference interval value, so as to obtain the difference value data set. If there is a condition greater than or equal to the preset difference threshold value in the difference value data set, such as the preset difference threshold value is 0.03, and the difference value between the calculated color similarity coefficient and the reference interval upper limit 0.95 is 0.05, the continue to check information is generated, which means that the connector needs to be further inspected.

[0033] According to the continue to check information, the appearance feature dynamic change data of the key detection area block and the surrounding auxiliary area block is collected, which specifically includes the following steps: The preset detection time length is set according to the detection time length requirement of the continue to check information. The appearance feature dynamic change data of the key detection area block and the surrounding auxiliary area block is collected at intervals within the preset detection time length, wherein the appearance feature dynamic change data includes color parameters, contour size deviation values and surface defect area proportion data.

[0034] The application first sets a preset detection time length according to the detection time length requirement in the continuous inspection information. If the detection time length required by the continuous inspection information is 10 minutes, the preset detection time length is set to 10 minutes. Within this preset 10-minute detection time length, the appearance feature dynamic change data of the key detection area block and the surrounding auxiliary area block is collected at intervals. The appearance feature dynamic change data includes color parameters, contour size deviation values and surface defect area proportion data. Taking a connector as an example, the key detection area block is the core contact part of the connector, and the surrounding auxiliary area block is the shell area around the core contact part. Within 10 minutes, the color parameters of the core contact part and the shell area are collected once every 1 minute, such as the color value of the core contact part metal and the color value of the shell plastic; at the same time, the contour size deviation values of them are collected, that is, the deviation of the actual measured size of the core contact part and the shell from the standard size; such as the surface defect area proportion data is the area proportion of the surface scratch of the core contact part and the area proportion of the small concave mark on the surface of the shell. Through this interval collection method, the dynamic change of the appearance feature with time is obtained.

[0035] The appearance feature dynamic change data is combined with the connector standard appearance parameter library to judge the first appearance state of the key detection area block, which specifically includes the following steps: A connector standard appearance parameter library is established. The connector standard appearance parameter library includes the standard color range, the standard size deviation threshold and the standard defect area proportion of the key detection area block of the connector under different working time lengths. The trend of the appearance feature dynamic change data is judged by a time series prediction model, and the first appearance state of the key detection area block after the preset detection time length is judged by combining the standard parameters corresponding to the time length in the connector standard appearance parameter library. The first appearance state includes color determination value, size deviation determination value and defect area proportion determination value.

[0036] The application needs to establish a connector standard appearance parameter library first. The connector standard appearance parameter library includes the standard color range, the standard size deviation threshold and the standard defect area proportion of the key detection area block of the connector under different working time lengths. For example, for a connector, the standard color range of the key detection area block is a specific color interval corresponding to the metal luster after working for 100 hours, the standard size deviation threshold is the allowed deviation range of the actual size from the standard size, and the standard defect area proportion is the proportion that the surface defect area of the region cannot exceed. When the working time length changes to 200 hours, these standard parameters will be adjusted according to historical data and actual conditions to adapt to the appearance change law of the connector in different working stages.

[0037] The trend of the dynamic change data of the appearance feature is judged by the time series prediction model. Taking the color parameter of the connector key detection area block as an example, assuming that the color parameter collected at intervals within the preset detection time shows a gradually deepening trend, the time series prediction model will obtain the rule and trend of color change based on these collected data. Then, combined with the standard parameters corresponding to the time length in the connector standard appearance parameter library, the first appearance state of the key detection area block after the preset detection time is judged. For example, after the preset detection time ends, the connector is expected to work for 150 hours, and the time series prediction model predicts the color determination value of the key detection area block at this time, and then compares it with the standard color range corresponding to 150 hours in the standard appearance parameter library; at the same time, the size deviation determination value and the defect area proportion determination value are predicted, and are compared with the standard size deviation threshold value and the standard defect area proportion respectively, so as to determine the first appearance state, including the color determination value, the size deviation determination value and the defect area proportion determination value.

[0038] The first appearance state of the key detection area block is corrected based on the feature correlation relationship between the key detection area block and the surrounding auxiliary area block to obtain a second appearance state; the second appearance state is processed to obtain a first quality risk value of the key detection area block, specifically including the following steps: The feature correlation weight between the key detection area block and the surrounding auxiliary area block is determined according to historical data, wherein the feature correlation weight includes color correlation weight, size correlation weight and defect correlation weight; The first appearance state of the surrounding auxiliary area block is multiplied by the corresponding feature correlation weight to obtain a weight correction value; The first appearance state of the key detection area block is corrected according to the weight correction value to obtain a second appearance state; An appearance state quality risk mapping model is established; The determination value, size deviation determination value and defect area proportion determination value of the second appearance state are input into the appearance state quality risk mapping model to obtain the first quality risk value of the key detection area block.

[0039] The historical data of the application contains the changes of various appearance features between the key detection area block and the peripheral auxiliary area block under different use scenarios and different working time lengths, as well as the corresponding quality results. For example, in the historical data research of a certain type of connector, it is found that when the color of the peripheral auxiliary area block changes obviously, there is an 80% probability that the color of the key detection area block will be abnormal, thereby affecting the electrical performance of the connector and causing problems such as poor contact. The color correlation weight is set to 0.7, the influence frequency and degree of the size change of the peripheral auxiliary area block on the size of the key detection area block, and the propagation rule of defects between the two areas are analyzed to determine the size correlation weight and the defect correlation weight respectively. Assuming that the size correlation weight is determined as 0.3 and the defect correlation weight is 0.5, these weight values reflect the importance of the correlation between different appearance features in the two area blocks.

[0040] The first appearance state of the peripheral auxiliary area block is multiplied by the corresponding feature correlation weight. Taking the color parameter as an example, if the color determination value in the first appearance state of the peripheral auxiliary area block is RGB(120, 120, 120) after measurement and calculation, and the color correlation weight is 0.7, then the weight correction value calculation method in terms of color is: for the RGB three channels, respectively, that is, (120x0.7, 120x0.7, 120x0.7), and the result is (84, 84, 84). The size and defect are also calculated according to the same logic. Assuming that the size deviation determination value in the first appearance state of the peripheral auxiliary area block is 2mm, and the size correlation weight is 0.3, then the weight correction value in terms of size is 2x0.3=0.6mm; if the defect area ratio determination value in the first appearance state of the peripheral auxiliary area block is 5%, and the defect correlation weight is 0.5, then the weight correction value in terms of defect is 5% x 0.5 = 2.5%, and the weight correction value of each feature is obtained through such calculation.

[0041] The first appearance state of the key detection area block is comprehensively adjusted according to the weight correction value. Taking color as an example, assuming that the color determination value of the key detection area block in the first appearance state is RGB(150, 150, 150), and combining the weight correction value of the color of the peripheral auxiliary area block (84, 84, 84), a more accurate color determination value of the second appearance state is obtained by weighted average. The specific calculation process is: RGB = ((150x0.3)+(84x0.7), (150x0.3)+(84x0.7), (150x0.3)+(84x0.7)) = (103.8, 103.8, 103.8). For the size deviation determination value and the defect area ratio determination value, the same is true according to the corresponding weight correction value, and the complete and accurate appearance parameter values in the second appearance state are obtained by adjusting according to the established correction algorithm.

[0042] A large amount of actual quality data of the connector in different appearance states is collected, including various cases from perfect appearance to various quality problems. Through statistical analysis and machine learning of these data, the internal relationship and rules between the appearance state parameters and the quality risk are mined. For example, it is found through analysis that when the color determination value of the key detection area block deviates from the standard value by a certain range, and the size deviation determination value and the defect area proportion determination value also reach certain thresholds at the same time, the connector has a serious quality problem. Based on these rules, an appearance state quality risk mapping model is constructed, which can accurately map each parameter of the appearance state to the corresponding quality risk value interval.

[0043] Finally, the determination value of the second appearance state, the size deviation determination value and the defect area proportion determination value are input into the appearance state quality risk mapping model. After receiving these parameters, the model calculates and judges according to the internal algorithm and rules. For example, it is judged whether the color determination value is within the normal range, and if not, the corresponding risk score is calculated according to the deviation degree; then the size deviation determination value and the defect area proportion determination value are processed in the same way. Then the risk scores of these three aspects are comprehensively calculated, and finally the first quality risk value of the key detection area block is output. This risk value is presented in the form of a specific numerical value, which directly reflects the degree of quality risk existing in the key detection area block, and provides an important basis for subsequent comprehensive evaluation and decision-making of the connector quality.

[0044] According to the environmental interference parameters of the corresponding assembly scene of the connector and the historical quality failure data, an environmental failure correlation model is constructed, and the first quality risk value is combined with the dynamic change data of the appearance feature to obtain the second quality risk value of the key detection area block through the environmental failure correlation model, which includes the following steps: The environmental interference parameters of the connector in the corresponding assembly scene are collected, wherein the environmental interference parameters include environmental temperature, humidity, dust concentration and vibration frequency; The historical quality failure data of the connector under the environmental interference parameters is collected, and an environmental failure database is established; The environmental failure database is trained by using a machine learning algorithm to obtain an environmental failure correlation model; The environmental interference parameters of the current assembly scene are input into the environmental failure correlation model to obtain an environmental influence coefficient, and the second quality risk value of the key detection area block is obtained according to the environmental influence coefficient and the first quality risk value.

[0045] The application first collects the environmental interference parameters of the connector in the corresponding assembly scene, which include environmental temperature, humidity, dust concentration and vibration frequency. Taking the connector assembly scene in an electronic equipment production workshop as an example, the staff continuously monitors the environmental temperature in the workshop using a professional temperature sensor; measures the humidity of the air with a hygrometer to understand whether it is in a relatively dry 40% relative humidity or the humidity reaches 60% and above due to water sources or air circulation in the workshop; detects the dust content in each cubic meter of air with the help of a dust concentration detector to determine whether the dust is in the normal low concentration range or the dust concentration is significantly increased due to cutting, grinding and other processes in the production process; and measures the vibration frequency generated by the operation of production equipment and the movement of transportation tools in the workshop to master the strength and frequency of vibration.

[0046] Collect historical quality failure data of the connector under these environmental interference parameters and establish an environmental failure database. For example, detailed records are kept of the specific conditions of quality failures such as poor contact, shell cracking, internal component damage, etc. of the connector when the workshop temperature is 30 degrees Celsius, the humidity is 60%, the dust concentration exceeds the standard, and the vibration frequency is a certain specific value, including the time of failure occurrence, the number of connectors involved, the type and severity of failure, etc. A large amount of historical data is classified, sorted and stored to build an environmental failure database.

[0047] An environmental failure correlation model is obtained by training the environmental failure database using a machine learning algorithm. The probability of connector contact failure is mined when the temperature rises to 32 degrees Celsius, the humidity exceeds 55%, the dust concentration per cubic meter reaches a certain value, and the vibration frequency is in a certain interval; different combinations of environmental parameters correspond to different types and probabilities of quality failure rules. Through continuous learning and training of the data in the environmental failure database, the model can accurately master these complex rules, thereby establishing an environmental failure correlation model that can accurately reflect the correlation between environmental interference parameters and quality failure.

[0048] The environmental interference parameters of the current assembly scene are input into the environmental failure correlation model to obtain an environmental impact coefficient. The environmental impact coefficient can quantify the degree of adverse impact of the current environment on the quality of the connector. The second quality risk value of the key detection area block is calculated according to the environmental impact coefficient and the first quality risk value. For example, if the first quality risk value itself indicates that the connector has a certain quality risk, and the environmental impact coefficient shows that the current environment will increase the quality risk by 30%, then the first quality risk value and the environmental impact coefficient are operated accordingly (such as the first quality risk value multiplied by 1.3), and the second quality risk value obtained is higher than the first quality risk value.

[0049] The comprehensive quality inspection result of the connector is obtained according to the first quality risk value and the second quality risk value, and specifically includes the following steps: The first risk threshold and the second risk threshold are preset, wherein the first risk threshold is less than the second risk threshold; If the first quality risk value and the second quality risk value are both lower than the first risk threshold, it is determined that the connector quality is qualified; If any of the first quality risk value and the second quality risk value is between the first risk threshold and the second risk threshold, it is determined that the connector needs to be repaired and re-inspected; If any of the first quality risk value and the second quality risk value is higher than the second risk threshold, it is determined that the connector quality is unqualified.

[0050] The first risk threshold and the second risk threshold are preset, wherein the first risk threshold is less than the second risk threshold.

[0051] The first quality risk value and the second quality risk value are compared with the two thresholds. If the first quality risk value and the second quality risk value are both lower than the first risk threshold, for example, the first quality risk value is 25 and the second quality risk value is 20, and the first quality risk value and the second quality risk value are both less than 30, it is determined that the connector quality is qualified.

[0052] If any of the first quality risk value and the second quality risk value is between the first risk threshold and the second risk threshold, for example, the first quality risk value is 40, between 30 and 60, and the second quality risk value is 25, lower than 30, it is determined that the connector needs to be repaired and re-inspected. This is because there is a quality risk value at the intermediate level, which means that the connector has potential quality problems and can only be determined whether it truly meets the quality requirements after repair.

[0053] If any of the first quality risk value and the second quality risk value is higher than the second risk threshold, for example, the first quality risk value is 70, greater than 60, it is determined that the connector quality is unqualified. This indicates that the connector has a serious quality risk and cannot be used normally, and needs to be scrapped or more thoroughly processed.

[0054] As a supplement, Figure 5As shown, it is a schematic diagram of multi-angle image acquisition and region division of the connector in one embodiment of the application. In this embodiment, the connector is placed on the detection platform, and images are synchronously or sequentially acquired from different illumination angles by multiple industrial cameras arranged around, which together constitute the multi-angle visual image set. After image acquisition is completed, processing is performed via the division module. As shown, the system automatically identifies and segments the key detection region block of the connector (for example, the region is highlighted as a grid, which usually corresponds to the metal contact terminal or core functional area of the connector) and the peripheral auxiliary region block (for example, the region is marked as a diagonal shadow, which usually corresponds to the plastic shell, buckle or solder pin of the connector, etc. Non-core area). This division process ensures that subsequent analysis can focus on the core quality point while taking into account the auxiliary influence of the surrounding structure.

[0055] As shown, it is a schematic diagram of multi-angle image acquisition and region division of the connector in one embodiment of the application. In this embodiment, the connector is placed on the detection platform, and images are synchronously or sequentially acquired from different illumination angles by multiple industrial cameras arranged around, which together constitute the multi-angle visual image set. After image acquisition is completed, processing is performed via the division module. As shown, the system automatically identifies and segments the key detection region block of the connector (for example, the region is highlighted as a grid, which usually corresponds to the metal contact terminal or core functional area of the connector) and the peripheral auxiliary region block (for example, the region is marked as a diagonal shadow, which usually corresponds to the plastic shell, buckle or solder pin of the connector, etc. Non-core area). This division process ensures that subsequent analysis can focus on the core quality point while taking into account the auxiliary influence of the surrounding structure. Figure 6 As shown, it is an integrated schematic diagram of environmental interference parameter acquisition and quality risk comprehensive determination in one embodiment of the application. The figure clearly shows the workflow of the second processing module in the actual assembly scene. The connector is placed on a simulated or real assembly environment (such as a circuit board). At this time, various environmental sensors deployed in the scene acquire environmental interference parameters in real time, including but not limited to: environmental temperature monitored by temperature sensors, air humidity monitored by humidity sensors, particulate matter content in the air monitored by dust concentration sensors, and device running vibration frequency monitored by vibration sensors. These real-time data are input into the pre-trained environmental failure correlation model. The model is built based on the historical quality failure database and can output a quantitative environmental influence coefficient. At the same time, the first quality risk value (calculated based on the appearance state) from the first processing module is also sent to the integrated system. Finally, the system comprehensively considers these two aspects of information, obtains a more comprehensive and accurate second quality risk value through a preset algorithm (such as weighted calculation), and transmits it to the output module to generate the final comprehensive quality inspection result, fully illustrating how the application combines the appearance state of the product itself with the actual use environment factors to realize static inspection to dynamic scene prediction.

[0056] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. can be located in one place or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0057] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0058] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A connector quality inspection system based on multi-angle vision, characterized in that, include: The segmentation module divides the connector into key detection area blocks and surrounding auxiliary area blocks by performing region segmentation on the multi-angle visual image set of the connector under different lighting angles. Analysis module: Processes and analyzes key detection area blocks and surrounding auxiliary area blocks to obtain quality anomaly warning information or further inspection information; Judgment Module: Based on the dynamic change data of the appearance features of the key inspection area block and the surrounding auxiliary area block collected by the continued inspection information, and combined with the connector standard appearance parameter library, the first appearance state of the key inspection area block is determined. First processing module: Based on the feature correlation between the key detection area block and the surrounding auxiliary area blocks, the first appearance state of the key detection area block is corrected to obtain the second appearance state; the second appearance state is processed to obtain the first quality risk value of the key detection area block; The second processing module constructs an environmental fault association model based on the environmental interference parameters of the assembly scenario corresponding to the connector and historical quality fault data. It then uses the environmental fault association model to obtain the second quality risk value of the key detection area block by combining the dynamic change data of appearance features and the first quality risk value. Output module: Based on the first quality risk value and the second quality risk value, a comprehensive quality inspection of the connector is performed to obtain the comprehensive quality inspection result.

2. The connector quality inspection system based on multi-angle vision according to claim 1, characterized in that, The key detection area and surrounding auxiliary area are processed and analyzed to obtain quality anomaly warning information or continued inspection information. This process includes the following steps: Extract the appearance features of key detection regions and surrounding auxiliary regions, as well as their feature correlation coefficients; The image data of the connector under historical conditions is processed and analyzed to obtain the feature association reference interval value; Within the current detection period, the set of current feature correlation coefficients between key detection area blocks and surrounding auxiliary area blocks is calculated. The current set of feature correlation coefficients is compared with the feature correlation reference interval value to obtain quality anomaly warning information or further inspection information.

3. The connector quality inspection system based on multi-angle vision according to claim 2, characterized in that, Extracting the appearance features and feature correlation coefficients between key detection regions and surrounding auxiliary regions involves the following steps: Extract the appearance feature information of the key detection area block and the surrounding auxiliary area block, wherein the appearance feature information includes color parameters, contour size deviation value and surface defect area ratio; Calculate the feature correlation coefficient set between the key detection area block and the surrounding auxiliary area block, wherein the feature correlation coefficient set includes color similarity coefficient, size correlation coefficient and defect influence coefficient.

4. The connector quality inspection system based on multi-angle vision according to claim 3, characterized in that, The image data of the connector under historical conditions is processed and analyzed to obtain the feature association reference interval value, which specifically includes the following steps: Extract multi-angle visual image data of the connector under historical normal working conditions and obtain a set of correlation coefficients for historical normal features; Extract multi-angle visual image data of connectors under historical quality failure states and obtain a set of correlation coefficients for historical failure features; The reference interval value for feature association is determined based on the distribution range of the historical normal feature association coefficient set and the abnormal threshold of the historical fault feature association coefficient set.

5. The connector quality inspection system based on multi-angle vision according to claim 4, characterized in that, The current set of feature correlation coefficients is compared with the feature correlation reference interval value to determine whether to obtain quality anomaly warning information or further inspection information. The specific steps include: The current set of feature correlation coefficients is compared with the corresponding feature correlation reference interval values. If the current feature correlation coefficient set exceeds the feature correlation reference interval value, then the region block corresponding to the correlation reference interval value is determined, and a quality anomaly warning message is generated. If the current feature association coefficient set is within the feature association reference interval, then the difference between the current feature association coefficient set and the upper limit of the corresponding feature association reference interval is calculated to obtain the difference dataset. If the difference dataset contains values ​​greater than or equal to a preset difference threshold, then further testing information is generated.

6. The connector quality inspection system based on multi-angle vision according to claim 5, characterized in that, Based on the dynamic changes in the appearance features of the key detection area and surrounding auxiliary area, the following steps are specifically included: Set the preset detection time according to the detection time requirements for continued inspection information; Dynamic change data of appearance features of key detection area blocks and surrounding auxiliary area blocks are collected at intervals within a preset detection time period. The dynamic change data of appearance features includes color parameters, contour size deviation values ​​and surface defect area ratio data.

7. A connector quality inspection system based on multi-angle vision according to claim 6, characterized in that, The first appearance state of key inspection areas is determined by combining dynamic change data of appearance features with a standard appearance parameter library for connectors. This process includes the following steps: Establish a standard appearance parameter library for connectors; wherein, the standard appearance parameter library for connectors includes the standard color range, standard size deviation threshold, and standard defect area ratio of key inspection area blocks for connectors under different working durations; The time series prediction model is used to determine the trend of dynamic changes in appearance features. Combined with the standard parameters of the corresponding duration in the connector standard appearance parameter library, the first appearance state of the key inspection area block after the preset inspection time is determined. The first appearance state includes the color determination value, the size deviation determination value, and the defect area ratio determination value.

8. A connector quality inspection system based on multi-angle vision according to claim 7, characterized in that, Based on the feature correlation between the key detection area block and the surrounding auxiliary area blocks, the first appearance state of the key detection area block is corrected to obtain the second appearance state; the second appearance state is then processed to obtain the first quality risk value of the key detection area block, specifically including the following steps: The feature association weights between the key detection area block and the surrounding auxiliary area block are determined based on historical data. The feature association weights include color association weights, size association weights, and defect association weights. The weight correction value is obtained by multiplying the first appearance state of the surrounding auxiliary area block with the corresponding feature association weight; The first appearance state of the key detection area block is corrected according to the weight correction value to obtain the second appearance state; Establish an appearance condition quality risk mapping model; The determined values ​​of the second appearance state, the dimensional deviation, and the defect area ratio are input into the appearance state quality risk mapping model to obtain the first quality risk value of the key inspection area block.

9. A connector quality inspection system based on multi-angle vision according to claim 8, characterized in that, An environmental fault correlation model is constructed based on environmental interference parameters and historical quality fault data of the connector assembly scenario. The second quality risk value of the key detection area block is obtained by combining dynamic change data of appearance features with the first quality risk value through the environmental fault correlation model. The specific steps include: Collect environmental interference parameters of the connector in the corresponding assembly scenario, wherein the environmental interference parameters include ambient temperature, humidity, dust concentration and vibration frequency; Collect historical quality failure data of connectors under environmental interference parameters and establish an environmental failure database; An environmental fault correlation model is obtained by training an environmental fault database using machine learning algorithms; The environmental interference parameters of the current assembly scenario are input into the environmental fault association model to obtain the environmental impact coefficient. Based on the environmental impact coefficient and the first quality risk value, the second quality risk value of the key detection area block is obtained.

10. A connector quality inspection system based on multi-angle vision according to claim 9, characterized in that, The connector is subjected to a comprehensive quality inspection based on the first quality risk value and the second quality risk value to obtain the comprehensive quality inspection result, which specifically includes the following steps: A primary risk threshold and a secondary risk threshold are preset, wherein the primary risk threshold is lower than the secondary risk threshold; If both the first quality risk value and the second quality risk value are lower than the first-level risk threshold, the connector is deemed to be of acceptable quality. If either the first quality risk value or the second quality risk value falls between the first-level risk threshold and the second-level risk threshold, the connector is determined to need repair and re-inspection. If either the first quality risk value or the second quality risk value is higher than the secondary risk threshold, the connector is deemed to be of substandard quality.

Images