Gluing quality detection method and system based on visual image

Abstract

The invention relates to the field of visual detection, in particular to a gluing quality detection method and system based on a visual image. The method comprises the following steps that S1, an industrial visual code scanner scans an identification code on a plate to obtain current plate information; s2, acquiring multi-angle gluing image data of the plate through industrial vision equipment; s3, based on the plate information and the multi-angle gluing image data, marking defect data through a gluing defect detection model, and determining gluing quality detection data of the plate; s4, based on the gluing quality detection data, performing sampling inspection on the detection data by a worker; and S5, feeding back data based on sampling inspection to a model training module, and performing intensive training on the gluing defect detection model. Defect detection is carried out through a visual detection mode of the visual equipment, so that the quality inspection result is determined, the quality inspection efficiency, quality and accuracy are improved, and the quality inspection timeliness is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of visual inspection, and particularly to a method and system for detecting the quality of glue application based on visual images. Background Art

[0002] With the development of industrial intelligent manufacturing, consumers have higher and higher requirements for the quality and appearance of wooden furniture. At the same time, the automated glue application and edge banding equipment in the furniture industry is complex, and there are problems such as unstable quality, which has brought a greater impact on wooden furniture manufacturing enterprises. In particular, how to effectively improve the efficiency and accuracy of quality inspection has become particularly important.

[0003] In the prior art, for the detection of the quality of glue application on panels, manual visual recognition and touch are usually added at the end of the production line of the edge banding and glue application equipment to detect the quality of glue application. However, the manual recognition method has problems such as low detection efficiency and poor detection accuracy. At the same time, the low timeliness of detection will result in a large amount of rework. Summary of the Invention

[0004] The purpose of the present invention is to address the problems in the background art and propose a method and system for detecting the quality of glue application based on visual images. Defect detection is performed through the visual detection method of a vision device to determine the quality inspection result, improving the efficiency, quality, and accuracy of quality inspection and enhancing the timeliness of quality inspection.

[0005] On the one hand, the present invention proposes a method for detecting the quality of glue application based on visual images, including the following steps:

[0006] S1. An industrial vision scanner scans the identification code on the panel to obtain the current panel information;

[0007] S2. Industrial vision devices are used to obtain multi-angle glue application image data of the panel;

[0008] S3. Based on the panel information and multi-angle glue application image data, defect data is marked through a glue application defect detection model to determine the glue application quality detection data of the panel;

[0009] S4. Based on the glue application quality detection data, workers conduct spot checks on the detection data;

[0010] S5. The data based on the spot checks is fed back to the model training module to perform reinforcement training on the glue application defect detection model.

[0011] Preferably, the identification code is a one-dimensional code or a two-dimensional code, and the panel information includes panel size, material, edge banding, and glue.

[0012] Preferably, light is projected from the side of the board piece through an industrial strip light source to reduce reflection. Multiple groups of industrial vision acquisition devices are placed above and below the side of the board piece edge gluing production line. The industrial vision acquisition devices are initialized and configured, and the acquisition frequency of the industrial vision acquisition devices is configured.

[0013] Preferably, S3 includes:

[0014] S31. According to the board piece size feature information, through an image stitching algorithm, the multi-angle gluing image data is fused into one image data;

[0015] S32. According to the board piece material, edge banding and glue feature information and the above-mentioned fused image data, defect detection is carried out through a gluing quality detection algorithm to determine the gluing quality detection data corresponding to the board piece. The gluing quality detection data includes the number of defects, defect positions, defect confidence levels, and defect image division data;

[0016] S33. After obtaining the board piece detection result, the detected data is stored corresponding to the board piece identification code.

[0017] Preferably, S4 includes:

[0018] S41. Based on the board piece identification code and the corresponding gluing quality detection data, the detection data information of the board piece is determined by scanning the code with a mobile phone;

[0019] S42. Based on the above-mentioned detection data information, the worker reports the manual detection result, and the reported data includes whether the defect is correct, the defect position, and the defect division range.

[0020] Preferably, S5 includes:

[0021] S51. The gluing defect detection model is strengthened and trained through a model training module. Among them, the training data includes 40% of the manually reported detection data for one month and 60% of the manually labeled data;

[0022] S52. The model trained by the model training module is applied to the gluing defect detection model.

[0023] On the other hand, the present invention proposes a glue application quality detection system based on visual images, which includes a panel information acquisition module, an image data acquisition module, a glue application quality detection module, an artificial sampling inspection module, and an enhanced training module; the panel information acquisition module uses an industrial vision scanner to scan the identification code on the panel to obtain the current panel information; the image data acquisition module obtains multi-angle glue application image data of the panel through industrial vision devices; the glue application quality detection module, based on the panel information and multi-angle glue application image data, marks the defect data through a glue application defect detection model to determine the glue application quality detection data of the panel; the artificial sampling inspection module, based on the glue application quality detection data, the worker conducts sampling inspection on the detection data; the enhanced training module feeds back the sampled data to the model training module to perform enhanced training on the glue application defect detection model.

[0024] Compared with the prior art, the present invention has the following beneficial technical effects:

[0025] The present invention respectively and real-time collects the panel information and multi-angle glue application image data through an industrial vision scanner and industrial vision devices, and based on the above data, conducts defect detection through a glue application defect detection model to determine the quality inspection result, improving the quality inspection efficiency, quality, and accuracy, and enhancing the timeliness of quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of the detection method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] Embodiment 1

[0028] As Figure 1 shown, a glue application quality detection method based on visual images proposed in this embodiment includes the following steps S1 - S5:

[0029] S1. The industrial vision scanner is arranged in front of the industrial vision acquisition device (relative to the production line direction), facing the identification code of the panel, and is connected to the detection system through a network cable. The industrial vision scanner scans the identification code on the panel to obtain the current panel information. The identification code is a one-dimensional code or a two-dimensional code, and the corresponding system stores and determines the information of the panel. The panel information includes panel size, material, edge banding, and glue.

[0030] S2. Obtain multi-angle glue application image data of the panel through industrial vision devices. The industrial bar light source is used to illuminate from the side of the panel to reduce reflection. Multiple groups of industrial vision acquisition devices are placed above and below the side of the panel edge gluing production line. Based on the light source conditions, initialize the configuration of the industrial vision acquisition devices, and configure the acquisition frequency of the industrial vision acquisition devices based on the production line speed.

[0031] S3. Based on the panel information and multi-angle glue application image data, use the glue application defect detection model to label the defect data and determine the glue application quality detection data of the panel, which specifically includes the following steps S31 - S33:

[0032] S31. According to the panel size feature information, use the image stitching algorithm to fuse the multi-angle glue application image data into one image data;

[0033] S32. According to the panel material, edge banding, and glue characteristics information and the above-fused image data, perform defect detection through the glue application quality detection algorithm to determine the corresponding glue application quality detection data of the panel. The glue application quality detection data includes the number of defects, defect positions, defect confidence levels, and defect image division data. Among them, the glue application defect detection model is an initial recognition model manually labeled and trained;

[0034] S33. After obtaining the panel detection results, store the detected data corresponding to the panel identification code into the system.

[0035] S4. Based on the glue application quality detection data, workers conduct spot checks on the detection data, which specifically includes the following steps S41 and S42:

[0036] S41. Based on the panel identification code and the corresponding glue application quality detection data, determine the panel detection data information by scanning the code with a mobile phone. Workers conduct spot checks on the defective and non-defective panels after quality inspection in batches;

[0037] S42. Based on the above detection data information, workers report the manual detection results to the system. The reported data includes whether the defect is correct, the defect position, and the defect division range.

[0038] S5. Feed back the data based on the spot checks to the model training module to perform reinforcement training on the glue application defect detection model, which specifically includes the following steps S51 and S52:

[0039] S51. Perform strengthening training on the glue application defect detection model through the model training module. Among them, the training data includes 40% of the manually reported detection data for one month and 60% of the manually labeled data, that is, the manually reported detection data and the manually labeled data account for 40% and 60% of the training data respectively;

[0040] S52. Apply the model trained by the model training module to the glue application defect detection model.

[0041] In this embodiment, an industrial vision scanner and an industrial vision device are used to respectively collect the information of the board to be quality inspected on the board edge gluing production line in real time and multi-angle gluing image data. Based on the above data, defect detection is carried out through a gluing defect detection model to determine the quality inspection result, which improves the quality inspection efficiency, quality and accuracy, and enhances the timeliness of quality inspection. In addition, the sampling data obtained by the manual sampling method is fed back to the model training module to enhance the training of the gluing defect detection model and improve the detection efficiency and accuracy of the gluing defect detection model.

[0042] Embodiment 2

[0043] This embodiment proposes a gluing quality detection system based on visual images for implementing the gluing quality detection method based on visual images in Embodiment 1. This detection system includes a board information acquisition module, an image data acquisition module, a gluing quality detection module, a manual sampling module, and an enhanced training module.

[0044] The board information acquisition module uses an industrial vision scanner to scan the identification code on the board to obtain the current board information.

[0045] The image data acquisition module obtains multi-angle gluing image data of the board through an industrial vision device.

[0046] The gluing quality detection module, based on the board information and multi-angle gluing image data, marks the defect data through a gluing defect detection model to determine the gluing quality detection data of the board.

[0047] Based on the gluing quality detection data, the manual sampling module allows workers to conduct sampling inspections on the detection data.

[0048] The enhanced training module feeds the data based on the sampling inspection back to the model training module to conduct enhanced training on the gluing defect detection model.

[0049] Each module in this embodiment is communicatively connected in sequence. The board information and multi-angle gluing image data are obtained visually, and the defect data is marked through a gluing defect detection model according to the obtained data to obtain the gluing quality detection data. The manual sampling data can also be fed back to the model training module to conduct enhanced training on the gluing defect detection model, improving the detection accuracy and detection efficiency of the gluing defect detection model.

[0050] The above has described the embodiments of the present invention in detail with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge scope of those skilled in the art to which the present invention pertains.

Claims

1. A method for detecting glue coating quality based on visual images, characterized in that: The following steps are involved: S1. The industrial visual barcode scanner scans the identification code on the board to obtain the current board information; S2. Obtain multi-angle gluing image data of panels through industrial visual equipment; S3, based on the panel information and multi-angle gluing image data, the defect data is marked by the gluing defect detection model to determine the gluing quality detection data of the panel; S4. Based on the glue coating quality test data, the workers conduct random inspections on the test data; S5. Feed the sampling-based data back to the model training module to conduct enhanced training on the glue coating defect detection model.

2. The method for detecting glue coating quality based on visual images according to claim 1, characterized in that: The identification code is a one-dimensional code or a two-dimensional code, and the panel information includes panel size, material, edge banding and glue.

3. The method for detecting glue coating quality based on visual images according to claim 2 is characterized in that: Use industrial strip light sources to illuminate the sides of the panels to reduce reflections, place multiple sets of industrial vision acquisition equipment on the upper and lower sides of the panel edge banding and gluing production line, initialize the industrial vision acquisition equipment, and configure the acquisition frequency of the industrial vision acquisition equipment.

4. The method for detecting the quality of glue coating based on visual images according to claim 3 is characterized in that S3 include: S31, according to the plate size feature information, through the image stitching algorithm, the multi-angle gluing image data is merged into one image data; S32, according to the material, edge band and glue feature information of the panel and the above-mentioned fused image data, defect detection is performed through a gluing quality detection algorithm to determine the gluing quality detection data corresponding to the panel, the gluing quality detection data including the number of defects, defect locations, defect confidence and defect image division data; S33, after obtaining the panel detection result, the detected data is stored corresponding to the panel identification code.

5. The method for detecting the quality of glue coating based on visual images according to claim 4, characterized in that S4 include: S41, based on the panel identification code and the corresponding gluing quality detection data, determine the detection data information of the panel by scanning the code with a mobile phone; S42. Based on the above detection data information, the worker reports the manual detection result, and the reported data includes whether the defect is correct, the defect location and the defect classification range.

6. The method for detecting the quality of glue coating based on visual images according to claim 5, characterized in that S5 include: S51, strengthening training of the glue coating defect detection model through a model training module, wherein the training data includes 40% of one month's manually reported detection data and 60% of manually labeled data; S52. Apply the model trained by the model training module to the gluing defect detection model.

7. A glue coating quality detection system based on visual images, used to implement the glue coating quality detection method based on visual images as claimed in claim 1, characterized in that: include: The panel information acquisition module uses an industrial visual barcode scanner to scan the identification code on the panel to obtain the current panel information; Image data acquisition module, which obtains multi-angle gluing image data of panels through industrial visual equipment; The gluing quality inspection module, based on the panel information and multi-angle gluing image data, marks the defect data through the gluing defect detection model to determine the gluing quality inspection data of the panel; Manual sampling module: based on the glue coating quality test data, workers conduct sampling inspections on the test data; The enhanced training module feeds back the sampling-based data to the model training module to perform enhanced training on the glue coating defect detection model.

Images