A product quality inspection method based on model guidance and visual analysis

Abstract

The present invention discloses a product quality detection method based on model guidance and visual analysis, which relates to the technical field of product quality detection and effectively improves the detection efficiency of product quality. The present invention establishes a textile coating dynamic model and a textile structure dynamic model based on optical video data and laser video data, and sets a plurality of dyeing monitoring points and textile monitoring points with the same position on the textile coating dynamic model and the textile structure dynamic model, and then establishes a monitoring range area according to each dyeing monitoring point and textile monitoring point, and generates a corresponding pixel value change curve and a normal position change interval, and maps the dyeing monitoring points and textile monitoring points of the textile coating dynamic model and the textile structure dynamic model onto the real-time coating dynamic model and the real-time structure dynamic model respectively, and then determines whether the corresponding textile product has an abnormality at the time, and repairs the corresponding textile product according to the abnormality result until the production of the corresponding textile product is completed.

Description

Technical Field

[0001] The present invention relates to the technical field of product quality detection, and in particular to a product quality detection method based on model guidance and visual analysis. Background Art

[0002] With the rapid development of science and technology, facing the increasingly complex production environment and market demand, the speed and variety of product production are constantly increasing, making product quality inspection work more limited and challenging. However, the existing technology has shortcomings in cost input, safety, and early warning accuracy. Therefore, it is necessary to introduce new technologies to improve the safety and accuracy of product quality early warning, so as to reduce the cost input of product quality inspection.

[0003] Therefore, how to improve the detection efficiency of product quality while ensuring the detection accuracy of product quality is a difficulty in the existing technology. To this end, a product quality detection method based on model guidance and visual analysis is provided. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a product quality detection method based on model guidance and visual analysis.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A product quality inspection method based on model guidance and visual analysis includes the following steps:

[0007] Step S1: setting a corresponding number of textile production stacks on a textile production line according to textile production steps, and installing a plurality of pairs of cameras and laser sensors on each textile production stack, so that when the textile production line produces textiles, optical video data and laser video data of the textile products are collected by the cameras and laser sensors;

[0008] Step S2: establishing a textile coating dynamic model and a textile structure dynamic model of the corresponding textile product based on the optical video data and the laser video data, and setting a plurality of dyeing monitoring points and textile monitoring points with the same position on the textile coating dynamic model and the textile structure dynamic model, respectively. Then, a monitoring range area is established based on each dyeing monitoring point and textile monitoring point, and a corresponding pixel value change curve and normal position change interval are generated;

[0009] Step S3: generating a real-time coating dynamic model and a real-time structural dynamic model of the corresponding textile product based on the optical video data and laser video data collected in real time, mapping the dyeing monitoring points and textile monitoring points of the textile coating dynamic model and the textile structural dynamic model onto the real-time coating dynamic model and the real-time structural dynamic model, respectively, to determine whether the corresponding textile product has an abnormality at that time;

[0010] Step S4: repair the corresponding textile product according to the abnormal result until the production of the corresponding textile product is completed.

[0011] Furthermore, the textile production steps are divided into weaving, pre-dyeing and dyeing, and each production step is composed of multiple production sub-steps. Then, according to the total number of production sub-steps of the textile production step, the same number of textile production stacks are set on the textile production line, and the conveyor belts between each textile production stack and each textile processing stack are numbered respectively;

[0012] The textile production stacks are equipped with twisting machines, dyeing machines, laser sensors, cameras and wireless communication devices, and the same cameras and laser sensors are set on the conveyor belts between the textile production stacks;

[0013] A processing life cycle is set for each type of textile product, and the processing life cycle is divided into several unit data collection cycles of equal length. The length of the unit data collection cycle is equal to the length of time each textile production stack uses to process the textile product.

[0014] Furthermore, the process of collecting the optical video data and the laser video data includes:

[0015] Each textile production stack on the textile production line produces raw materials for processing in turn, while optical video data and laser video data of the processing process are collected through laser sensors and cameras;

[0016] Before the textile processing products enter the next textile production stack, the camera and laser sensor located on the conveyor belt collect optical image data and laser image data of the textile processing products;

[0017] By collecting the optical video data, laser video data, optical image data and laser image data, real-time processing record data and several copies of historical processing record data of various types of textile products are collected;

[0018] The processing record data includes optical video data, laser video data, optical image data and laser image data, and each data is marked with the number of the corresponding textile processing stack or conveyor belt.

[0019] Furthermore, the process of establishing the dynamic model of the textile coating includes:

[0020] Dividing the optical video data of each production sub-step into a number of optical image data by frame, and extracting the textile product part from each optical image data through an image feature extraction process;

[0021] Obtaining the pixel value of each pixel in the textile product part, establishing a two-dimensional coordinate system, and overlapping and mapping the optical image data of the same textile type on the two-dimensional coordinate system in chronological order;

[0022] According to the unit data acquisition cycle length, k time nodes are set, as well as pixel change thresholds and sample quantity thresholds. If there is a pixel point position in the historical processing record data that is greater than or equal to the sample quantity threshold and in the continuous optical image data corresponding to the optical video data, and the pixel value change value is greater than or equal to the sample quantity threshold, then the corresponding pixel point position is marked as a staining monitoring point. Otherwise, no operation is performed, where k is a natural number greater than 3.

[0023] Repeat the above process of finding dyeing monitoring points, and then establish a corresponding textile coating static model based on the textile product part in each optical image data. Then, according to the time division order of the optical image data, the various textile coating static models are spliced ​​in sequence to obtain a textile coating dynamic model. Several coating model areas of the same size are divided in the textile coating dynamic model, and each dyeing monitoring point is marked on the corresponding coating model area in the textile coating dynamic model.

[0024] Furthermore, the process of establishing the textile structure dynamic model includes:

[0025] Establishing a corresponding textile structure dynamic model based on historical laser video data of production sub-steps, and dividing a number of structural model regions in the textile structure dynamic model that are identical to the coating model regions in the textile coating dynamic model;

[0026] Then, the dyeing monitoring points on the textile coating dynamic model are mapped onto the textile structure dynamic model, and the mapping positions are recorded as textile monitoring points. The dyeing monitoring points and textile monitoring points on the textile coating dynamic model and the textile structure dynamic model generated in each unit data collection cycle are mapped onto the textile coating dynamic model and the textile structure dynamic model at the beginning of the unit data collection cycle.

[0027] Furthermore, the process of establishing the monitoring range area includes:

[0028] Setting a monitoring range area, and taking each textile monitoring point and dyeing monitoring point as the center position of the monitoring range area, so that the area composed of all monitoring range areas covers the entire textile structure dynamic model and textile coating dynamic model;

[0029] The starting time node of the unit data acquisition cycle is selected as the initial time node, and then the pixel value change curve of each coating model area in each monitoring range area is recorded from the initial time node, as well as the position change trajectory and appearance time node of each structural model area in each monitoring range area;

[0030] Establish a three-dimensional coordinate system, generate position change trajectories for the laser video data collected during the same time sequence but in different processing life cycles of a type of textile product, and overlay and map them onto the same three-dimensional coordinate system. Then, divide the position change trajectories into several position change points, and perform a normal distribution on the position change points at each time node. Then, select the 25% left and right positions of the normal distribution center as normal position change points.

[0031] The normal position change points at each time point are sequentially spliced ​​together to obtain the normal position change intervals of the corresponding type of textile products in each production sub-step;

[0032] At the same time, the process of establishing the normal position change interval is adopted to obtain the normal occurrence time interval of each structural model area in the corresponding production sub-cycle;

[0033] Then, the normal position change interval and normal occurrence time interval of each production sub-step are mapped onto the corresponding textile structure dynamic model.

[0034] Furthermore, the process of determining whether the corresponding textile product is abnormal at the time includes:

[0035] generating a real-time structural dynamic model and a real-time coating dynamic model of the corresponding textile product based on the real-time optical video data and the real-time laser video data, and then, based on the unit data acquisition period corresponding to the real-time structural dynamic model and the real-time coating dynamic model and the type of textile product, retrieving the corresponding textile structural dynamic model and the textile coating dynamic model for overlapping mapping with the real-time structural dynamic model and the real-time coating dynamic model;

[0036] Then, according to the pixel value change curve of each coating model area in each monitoring range area on the textile coating dynamic model, it is judged whether the pixel value change of each real-time coating model area at each time node in the real-time coating dynamic model is normal;

[0037] At the same time, according to the normal position change interval and normal occurrence time interval of each structural model area in each monitoring range area on the textile structural dynamic model, it is judged whether there is an abnormality in each real-time structural model area in the corresponding real-time structural dynamic model;

[0038] The textile structure repair decision is generated based on the judgment result and executed in the next textile production stack.

[0039] Furthermore, the process of repairing the textile product according to the abnormal result includes:

[0040] When the wireless communication device in each textile production stack receives a textile coating repair decision or a textile structure repair decision, at the beginning of the next unit data collection cycle, the wireless communication device locates the abnormal position of the textile product according to the textile coating repair decision or the textile structure repair decision, and then calls the dyeing machine or the twisting machine to repair the corresponding position. After the repair is completed, the production sub-step is executed to process the textile product;

[0041] Repeat steps S3 and S4 until all production sub-steps of the corresponding type of textile product are completed.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention establishes a textile coating dynamic model and a textile structure dynamic model based on optical video data and laser video data, and sets a number of dyeing monitoring points and textile monitoring points with the same position on the textile coating dynamic model and the textile structure dynamic model, and then establishes a monitoring range area according to each dyeing monitoring point and textile monitoring point, and generates corresponding pixel value change curves and normal position change intervals, and maps the dyeing monitoring points and textile monitoring points of the textile coating dynamic model and the textile structure dynamic model onto the real-time coating dynamic model and the real-time structure dynamic model respectively, and then determines whether the corresponding textile product has an abnormality at that time, and repairs the corresponding textile product according to the abnormal result until the production of the corresponding textile product is completed, thereby achieving the goal of improving the detection efficiency of product quality while ensuring the detection accuracy of product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention.

[0045] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention. Example

[0047] like Figure 1 As shown, a product quality inspection method based on model guidance and visual analysis includes the following steps:

[0048] Step S1: setting a corresponding number of textile production stacks on a textile production line according to textile production steps, and installing a plurality of pairs of cameras and laser sensors on each textile production stack, so that when the textile production line produces textiles, optical video data and laser video data of the textile products are collected by the cameras and laser sensors;

[0049] Step S2: establishing a textile coating dynamic model and a textile structure dynamic model of the corresponding textile product based on the optical video data and the laser video data, and setting a plurality of dyeing monitoring points and textile monitoring points with the same position on the textile coating dynamic model and the textile structure dynamic model, respectively. Then, a monitoring range area is established based on each dyeing monitoring point and textile monitoring point, and a corresponding pixel value change curve and normal position change interval are generated;

[0050] Step S3: generating a real-time coating dynamic model and a real-time structural dynamic model of the corresponding textile product based on the optical video data and laser video data collected in real time, mapping the dyeing monitoring points and textile monitoring points of the textile coating dynamic model and the textile structural dynamic model onto the real-time coating dynamic model and the real-time structural dynamic model, respectively, to determine whether the corresponding textile product has an abnormality at that time;

[0051] Step S4: repair the corresponding textile product according to the abnormal result until the production of the corresponding textile product is completed. Example

[0052] This embodiment further limits the embodiment 1, and the step S1 is implemented by the following process:

[0053] The textile production steps are divided into weaving, pre-dyeing and dyeing, and each production step is composed of multiple production sub-steps. Then, according to the total number of production sub-steps of the textile production step, the same number of textile production stacks are set on the textile production line, and the conveyor belts between each textile production stack and each textile processing stack are respectively numbered a. 1，1 、a 1，2 、……、a n，m , b1, b2, ..., b n*m-1 Among them, number a n，m represents the textile production stack corresponding to the m-th production sub-step of the n-th production step, where n and m are natural numbers greater than 0;

[0054] The textile production stacks are equipped with twisting machines, dyeing machines, laser sensors, cameras and wireless communication devices, and the same cameras and laser sensors are set on the conveyor belts between the textile production stacks;

[0055] A processing life cycle is set for each type of textile product, and the processing life cycle is divided into several unit data collection cycles of equal length. The length of the unit data collection cycle is equal to the length of time each textile production stack uses to process the textile product.

[0056] Furthermore, before textile production begins, workers upload the type and name of the textile product to be produced to a wireless communication device. Each textile production stack on the textile production line then processes the raw materials in turn, while simultaneously collecting optical and laser video data of the processing process through laser sensors and cameras.

[0057] Before the textile processing products enter the next textile production stack, the camera and laser sensor located on the conveyor belt collect optical image data and laser image data of the textile processing products;

[0058] By collecting the optical video data, laser video data, optical image data and laser image data, real-time processing record data and several copies of historical processing record data of various types of textile products are collected;

[0059] The processing record data includes optical video data, laser video data, optical image data and laser image data, and each data is marked with the number of the corresponding textile processing stack or conveyor belt. Example

[0060] This embodiment further limits the embodiment 1, and the step S2 is implemented by the following process:

[0061] The optical video data and laser video data in the historical processing record data of each type of textile product are arranged in sequence according to the subscript numbers of the numbers, and the textile coating dynamic model and textile structure dynamic model corresponding to the production sub-step are respectively established based on the optical video data and the laser video data;

[0062] The process of establishing the dynamic model of textile coating includes:

[0063] Dividing the optical video data of each production sub-step into a number of optical image data by frame, and extracting the textile product part from each optical image data through an image feature extraction process;

[0064] Obtaining the pixel value of each pixel in the textile product part, establishing a two-dimensional coordinate system, and overlapping and mapping the optical image data of the same textile type on the two-dimensional coordinate system in chronological order;

[0065] According to the unit data acquisition cycle length, k time nodes are set, as well as pixel change thresholds and sample quantity thresholds. If there is a pixel point position in the historical processing record data that is greater than or equal to the sample quantity threshold and in the continuous optical image data corresponding to the optical video data, and the pixel value change value is greater than or equal to the sample quantity threshold, then the corresponding pixel point position is marked as a staining monitoring point. Otherwise, no operation is performed, where k is a natural number greater than 3.

[0066] Repeat the above process of finding dyeing monitoring points, and then establish a corresponding textile coating static model based on the textile product part in each optical image data. Then, according to the time division order of the optical image data, the various textile coating static models are spliced ​​in sequence to obtain a textile coating dynamic model. Several coating model areas of the same size are divided in the textile coating dynamic model, and each dyeing monitoring point is marked on the corresponding coating model area in the textile coating dynamic model.

[0067] Furthermore, the process of establishing the dynamic model of the textile structure includes:

[0068] Since laser video data is generated by a laser sensor sending a laser signal to a textile product, the laser video data can accurately record the spinning process of a twisting machine through the production of raw materials. Then, a corresponding textile structure dynamic model is established based on the historical laser video data of the production sub-steps, and several structural model areas are divided in the textile structure dynamic model that are the same as the coating model areas in the textile coating dynamic model.

[0069] Since the historical optical video data and historical laser video data corresponding to the textile coating dynamic model and the textile structure dynamic model are generated for the same textile product at the same time, the textile coating dynamic model and the textile structure dynamic model have the same components at each time node;

[0070] Then, the dyeing monitoring points on the textile coating dynamic model are mapped onto the textile structure dynamic model, and the mapped positions are recorded as textile monitoring points;

[0071] Since there are situations where textile monitoring points or dyeing monitoring points do not appear at the beginning of each unit data collection cycle, the dyeing monitoring points and textile monitoring points on the textile coating dynamic model and textile structure dynamic model generated by each unit data collection cycle are mapped to the textile coating dynamic model and textile structure dynamic model at the beginning of the corresponding unit data collection cycle.

[0072] Furthermore, a monitoring range area is set, and each textile monitoring point and dyeing monitoring point is used as the center position of the monitoring range area, so that the area composed of all monitoring range areas covers the entire textile structure dynamic model and the textile coating dynamic model;

[0073] It should be noted that if there is an area composed of all monitoring range areas that cannot cover the entire textile structure dynamic model and textile coating dynamic model, the uncovered area will be directly recorded as the monitoring range area;

[0074] The starting time node of the unit data acquisition cycle is selected as the initial time node, and then the pixel value change curve of each coating model area in each monitoring range area is recorded from the initial time node, as well as the position change trajectory and appearance time node of each structural model area in each monitoring range area;

[0075] Establish a three-dimensional coordinate system, generate position change trajectories for the laser video data collected during the same time sequence but in different processing life cycles of a type of textile product, and overlay and map them onto the same three-dimensional coordinate system. Then, divide the position change trajectories into several position change points, and perform a normal distribution on the position change points at each time node. Then, select the 25% left and right positions of the normal distribution center as normal position change points.

[0076] The normal position change points at each time point are sequentially spliced ​​together to obtain the normal position change intervals of the corresponding type of textile products in each production sub-step;

[0077] At the same time, the process of establishing the normal position change interval is adopted to obtain the normal occurrence time interval of each structural model area in the corresponding production sub-cycle;

[0078] It should be noted that for the structural model regions that exist at the beginning of the unit data collection period, their normal occurrence time interval is not recorded;

[0079] Then, the normal position change interval and normal occurrence time interval of each production sub-step are mapped onto the corresponding textile structure dynamic model. Example

[0080] This embodiment further limits the embodiment 1, and the step S3 is implemented by the following process:

[0081] generating a real-time structural dynamic model and a real-time coating dynamic model of the corresponding textile product based on the real-time optical video data and the real-time laser video data, and then, based on the unit data acquisition period corresponding to the real-time structural dynamic model and the real-time coating dynamic model and the type of textile product, retrieving the corresponding textile structural dynamic model and the textile coating dynamic model for overlapping mapping with the real-time structural dynamic model and the real-time coating dynamic model;

[0082] Then, according to the pixel value change curve of each coating model area in each monitoring range area on the textile coating dynamic model, it is judged whether the pixel value change of each real-time coating model area at each time node in the real-time coating dynamic model is normal;

[0083] Set a pixel edge threshold. If the difference between the pixel change value of the real-time coating model area and the corresponding value of the pixel value change curve is greater than or equal to the pixel edge threshold, the pixel value change of the corresponding real-time coating model area is judged to be abnormal. Otherwise, it is judged to be normal and the coating judgment of the next production sub-step is carried out;

[0084] If the number of the real-time optical video data corresponding to the textile structure dynamic model where the pixel value change is abnormal is determined to be a conveyor belt number, a textile coating repair decision is generated based on the location where the pixel value change is abnormal and is executed at the next textile production stack;

[0085] At the same time, according to the normal position change interval and normal occurrence time interval of each structural model area in each monitoring range area on the textile structural dynamic model, it is judged whether there is an abnormality in each real-time structural model area in the corresponding real-time structural dynamic model;

[0086] If the real-time position change trajectory and appearance time of each real-time structural model area in the real-time coating dynamic model are within the normal position change interval and normal appearance time interval, then the current production sub-step is judged to be normal;

[0087] Otherwise, it is determined that there is a structural anomaly in the corresponding structural model area, and then a textile structure repair decision is generated according to the location where the structural anomaly occurs, and it is executed in the next textile production stack. Example

[0088] This embodiment further limits the embodiment 1, and the step S4 is implemented by the following process:

[0089] When the wireless communication device in each textile production stack receives a textile coating repair decision or a textile structure repair decision, at the beginning of the next unit data collection cycle, the wireless communication device locates the abnormal position of the textile product according to the textile coating repair decision or the textile structure repair decision, and then calls the dyeing machine or the twisting machine to repair the corresponding position. After the repair is completed, the production sub-step is executed to process the textile product;

[0090] It should be noted that when the textile production stack receives both the textile coating repair decision and the textile structure repair decision, the textile structure repair decision is executed first. When all textile structure repair decisions are completed, the textile coating repair decision is executed.

[0091] Repeat steps S3 and S4 until all production sub-steps of the corresponding type of textile product are completed.

[0092] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A product quality inspection method based on model guidance and visual analysis, characterized in that: The following steps are involved: Step S1: setting a corresponding number of textile production stacks on a textile production line according to textile production steps, and installing a plurality of pairs of cameras and laser sensors on each textile production stack, so that when the textile production line produces textiles, optical video data and laser video data of the textile products are collected by the cameras and laser sensors; Step S2: establishing a textile coating dynamic model and a textile structure dynamic model of the corresponding textile product based on the optical video data and the laser video data, and setting a plurality of dyeing monitoring points and textile monitoring points with the same position on the textile coating dynamic model and the textile structure dynamic model, respectively. Then, a monitoring range area is established based on each dyeing monitoring point and textile monitoring point, and a corresponding pixel value change curve and normal position change interval are generated; Step S3: generating a real-time coating dynamic model and a real-time structural dynamic model of the corresponding textile product based on the optical video data and laser video data collected in real time, mapping the dyeing monitoring points and textile monitoring points of the textile coating dynamic model and the textile structural dynamic model onto the real-time coating dynamic model and the real-time structural dynamic model, respectively, to determine whether the corresponding textile product has an abnormality at that time; Step S4: repairing the corresponding textile product according to the abnormal result until the production of the corresponding textile product is completed; The process of establishing the dynamic model of the textile coating includes: Dividing the optical video data of each production sub-step into a number of optical image data by frame, and extracting the textile product part from each optical image data through an image feature extraction process; Obtaining the pixel value of each pixel in the textile product part, establishing a two-dimensional coordinate system, and overlapping and mapping the optical image data of the same textile type on the two-dimensional coordinate system in chronological order; Set the pixel change threshold and sample number threshold. If a pixel point exists in the historical processing record data that is greater than or equal to the sample number threshold and the continuous optical image data corresponding to the optical video data has a pixel value change greater than or equal to the sample number threshold, then the corresponding pixel point is marked as a staining monitoring point. Otherwise, no operation is performed. Repeat the above process of finding the dyeing monitoring points, and then establish the corresponding textile coating static model according to the textile product part in each optical image data. Then, according to the time division order of the optical image data, the textile coating static models are sequentially spliced ​​to obtain the textile coating dynamic model. Several coating model areas of equal size are divided in the textile coating dynamic model, and each dyeing monitoring point is marked on the corresponding coating model area in the textile coating dynamic model; The process of establishing the monitoring range area includes: Set the monitoring range area, and use each textile monitoring point and dyeing monitoring point as the center position of the monitoring range area, select the starting time node of the unit data collection cycle as the initial time node, and then record the pixel value change curve of each coating model area in each monitoring range area from the initial time node, and record the position change trajectory and appearance time node of each structural model area in each monitoring range area; Establish a three-dimensional coordinate system, overlap and map the position change trajectories generated by laser video data collected during unit data collection cycles of different processing life cycles of a type of textile product but in the same time sequence onto the same three-dimensional coordinate system, divide the position change trajectories into several position change points, and perform a normal distribution on the position change points at each time node. Based on the selected normal distribution results, obtain the normal position change range of the corresponding type of textile product in each production sub-step; At the same time, the process of establishing the normal position change interval is adopted to obtain the normal occurrence time interval of each structural model area in the corresponding production sub-cycle; The normal position change interval and normal occurrence time interval of each production sub-step are mapped onto the corresponding textile structure dynamic model.

2. A product quality inspection method based on model guidance and visual analysis according to claim 1, characterized in that: The textile production steps are divided into weaving, pre-dyeing and dyeing, each production step is composed of multiple production sub-steps, and the same number of textile production stacks are set on the textile production line according to the total number of production sub-steps of the textile production step, and the conveyor belts between each textile production stack and each textile processing stack are respectively numbered; The textile production stacks are equipped with twisting machines, dyeing machines, laser sensors, cameras and wireless communication devices, and the same cameras and laser sensors are set on the conveyor belts between the textile production stacks; A processing life cycle is set for each type of textile product, and the processing life cycle is divided into several unit data collection cycles of equal length. The length of the unit data collection cycle is equal to the length of time each textile production stack uses to process the textile product.

3. The product quality inspection method based on model guidance and visual analysis according to claim 2, characterized in that: The acquisition process of the optical video data and the laser video data includes: Each textile production stack on the textile production line produces raw materials for processing in turn, while optical video data and laser video data of the processing process are collected through laser sensors and cameras; Before the textile processing products enter the next textile production stack, the camera and laser sensor located on the conveyor belt collect optical image data and laser image data of the textile processing products; By collecting the optical video data, laser video data, optical image data and laser image data, the real-time processing record data and several copies of historical processing record data of various types of textile products are collected.

4. The product quality inspection method based on model guidance and visual analysis according to claim 3 is characterized in that: The process of establishing the textile structure dynamic model includes: A textile structure dynamic model is established based on historical laser video data of production sub-steps, and a number of structural model regions that are identical to coating model regions in the textile coating dynamic model are divided in the textile structure dynamic model; Then, the dyeing monitoring points on the textile coating dynamic model are mapped onto the textile structure dynamic model, and the mapping positions are recorded as textile monitoring points. The dyeing monitoring points and textile monitoring points on the textile coating dynamic model and the textile structure dynamic model generated by each unit data collection cycle are mapped onto the textile coating dynamic model and the textile structure dynamic model at the beginning of the corresponding unit data collection cycle.

5. The product quality inspection method based on model guidance and visual analysis according to claim 4 is characterized in that: The process of determining whether the corresponding textile product has an abnormality at the time includes: Generate a real-time structural dynamic model and a real-time coating dynamic model of the corresponding textile product according to the real-time optical video data and the real-time laser video data, and retrieve the corresponding textile structural dynamic model and the textile coating dynamic model for overlapping mapping with the real-time structural dynamic model and the real-time coating dynamic model; Then, according to the pixel value change curve of each coating model area in each monitoring range area on the textile coating dynamic model, it is judged whether the pixel value change of each real-time coating model area at each time node in the real-time coating dynamic model is normal; At the same time, according to the normal position change interval and normal occurrence time interval of each structural model area in each monitoring range area on the textile structural dynamic model, it is judged whether there is an abnormality in each real-time structural model area in the corresponding real-time structural dynamic model; The textile structure repair decision is generated based on the judgment result and executed in the next textile production stack.

6. The product quality inspection method based on model guidance and visual analysis according to claim 5, characterized in that: The process of repairing textile products based on abnormal results includes: When the wireless communication device in each textile production stack receives a textile coating repair decision or a textile structure repair decision, at the beginning of the next unit data collection cycle, the wireless communication device locates the abnormal position of the textile product according to the textile coating repair decision or the textile structure repair decision, and then calls the dyeing machine or the twisting machine to repair the corresponding position. After the repair is completed, the production sub-step is executed to process the textile product; Repeat steps S3 and S4 until all production sub-steps of the corresponding type of textile product are completed.

Images