Gypsum board defect detection method and system based on unidirectional structured light

Abstract

The invention relates to the technical field of plasterboard production, in particular to a plasterboard defect detection method and system based on unidirectional structured light, and the method comprises the following steps: emitting unidirectional structured light for detecting plasterboard defects from one side of a plasterboard conveying assembly to the surface of a plasterboard at a preset angle by using a structured light emitter; an industrial camera is used for collecting a detection image formed by irradiation of the one-way structured light on the surface of the gypsum board and used for detecting the defects of the gypsum board; and performing defect identification on the detection image by using a multi-classifier to obtain a defect detection result of the detection image. The detection image for detecting the defects of the gypsum board is formed by utilizing the unidirectional structured light, the defect identification is performed on the detection image by utilizing the multiple classifiers, the defect detection result of the detection image is obtained, the data processing amount is small, and the detection cost of the unidirectional structured light is low.

Description

Technical Field

[0001] The present invention relates to the technical field of gypsum board production, and in particular to a gypsum board defect detection method and system based on unidirectional structured light. Background Art

[0002] Gypsum board is a material made primarily from building gypsum. It boasts lightweight, high strength, thin thickness, easy processing, and excellent sound insulation, heat insulation, and fire resistance. However, during the production process, gypsum board can often develop various defects due to various factors, including equipment, processes, and raw materials. Therefore, rigorous quality testing is essential before any product is shipped.

[0003] Existing gypsum board defect detection methods mostly determine the surface uniformity of the gypsum board by measuring its thickness or width. However, this method requires the deployment of a large number of laser detection systems on the inspection surface of the gypsum board, which increases costs and data processing, increasing the burden on system operation. Summary of the Invention

[0004] The purpose of the present invention is to provide a gypsum board defect detection method based on unidirectional structured light to solve the technical problems in the prior art of increased cost, excessive data processing, and increased system operation burden.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] A gypsum board defect detection method based on unidirectional structured light comprises the following steps:

[0007] Using a structured light emitter to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly at a preset angle toward the gypsum board surface;

[0008] Using an industrial camera to collect a detection image for detecting defects in the gypsum board formed by irradiating the unidirectional structured light on the surface of the gypsum board;

[0009] Using multiple classifiers to perform defect recognition on the detection image to obtain a defect detection result of the detection image;

[0010] According to the defect detection result of the detection image, a corresponding defect detection result of the gypsum board is obtained.

[0011] As a preferred solution of the present invention, the detection image is preprocessed into a uniform specification before defect recognition.

[0012] As a preferred solution of the present invention, the method of using multiple classifiers to perform defect recognition on the detection image to obtain the defect detection result of the detection image includes:

[0013] Inputting the detection image into a multi-classifier, and having the multi-classifier output a defect detection label of the detection image;

[0014] The defect detection labels include gypsum board defect labels and gypsum board non-defect labels, and the gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels;

[0015] The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively;

[0016] The gypsum board protrusion defect label corresponds to the detection result of the gypsum board having a protrusion defect and the detection result of the gypsum board having a concave defect.

[0017] As a preferred embodiment of the present invention, the training of the multiple classifiers includes:

[0018] Obtain a set of detection images as training samples;

[0019] Label a set of training samples to obtain the defect detection label of each training sample;

[0020] The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier;

[0021] Using a softmax classifier to perform classification learning on the input items and the output items to obtain a multi-classifier for defect recognition of the inspection image;

[0022] The multi-classifier is:

[0023] Label = softmax(G);

[0024] Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

[0025] As a preferred solution of the present invention, all training samples are normalized.

[0026] As a preferred embodiment of the present invention, the present invention provides a gypsum board defect detection system based on unidirectional structured light, which is applied to the gypsum board defect detection method based on unidirectional structured light. The gypsum board defect detection system includes:

[0027] a structured light emitter, configured to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly toward the gypsum board surface at a preset angle;

[0028] an industrial camera for collecting an inspection image formed by irradiating the gypsum board surface with the unidirectional structured light for detecting defects in the gypsum board;

[0029] A data processor, configured to perform defect recognition on the detection image using a multi-classifier to obtain a defect detection result of the detection image;

[0030] The detection output portal is used to obtain the defect detection result of the gypsum board according to the defect detection result of the detection image.

[0031] As a preferred solution of the present invention, the industrial camera is located below the surface of the gypsum board and vertically captures the detection image formed by the unidirectional structured light irradiated on the surface of the gypsum board.

[0032] As a preferred solution of the present invention, the data processor uses multiple classifiers to perform defect recognition on the detection image to obtain a defect detection result of the detection image, including:

[0033] Inputting the detection image into a multi-classifier, and having the multi-classifier output a defect detection label of the detection image;

[0034] The defect detection labels include gypsum board defect labels and gypsum board non-defect labels, and the gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels;

[0035] The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively;

[0036] The gypsum board protrusion defect label corresponds to the detection result of the gypsum board having a protrusion defect and the detection result of the gypsum board having a concave defect.

[0037] As a preferred solution of the present invention, the data processor trains multiple classifiers, including:

[0038] Obtain a set of detection images as training samples;

[0039] Label a set of training samples to obtain the defect detection label of each training sample;

[0040] The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier;

[0041] Using a softmax classifier to perform classification learning on the input items and the output items to obtain a multi-classifier for defect recognition of the inspection image;

[0042] The multi-classifier is:

[0043] Label = softmax(G);

[0044] Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

[0045] As a preferred solution of the present invention, the data processor preprocesses the detection images into a uniform specification and performs normalization processing on the training samples.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention uses unidirectional structured light to form a detection image for detecting gypsum board defects, uses multiple classifiers to identify defects in the detection image, and obtains a defect detection result of the detection image. The data processing volume is small, and the detection cost of the unidirectional structured light is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0049] Figure 1 A flow chart of a gypsum board defect detection method based on unidirectional structured light provided in an embodiment of the present invention;

[0050] Figure 2 A block diagram of a gypsum board defect detection system provided by an embodiment of the present invention;

[0051] Figure 3 This is a detection image provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, the present invention provides a gypsum board defect detection method based on unidirectional structured light, comprising the following steps:

[0054] Using a structured light emitter to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly at a preset angle toward the gypsum board surface;

[0055] An industrial camera is used to collect images formed by irradiating the gypsum board surface with unidirectional structured light to detect defects in the gypsum board, such as Figure 3 As shown, the gray lines are unidirectional structured light;

[0056] Use multiple classifiers to identify defects in the inspection image and obtain defect detection results of the inspection image;

[0057] According to the defect detection result of the detection image, the corresponding defect detection result of the gypsum board is obtained.

[0058] In order to reduce detection costs and improve detection efficiency, the present invention forms unidirectional structured light on the surface of the gypsum board, uses a high-speed camera to shoot, and determines whether there are defects based on the degree of light bending of the structured light, thereby avoiding the need to arrange a large number of laser detection systems on the detection surface of the gypsum board, achieving cost reduction, reducing the amount of data processing, enhancing defect detection efficiency, and meeting the timeliness requirements of defect detection.

[0059] Furthermore, the present invention is combined with a classifier model in deep learning during unidirectional structured light defect detection to identify whether there are defects on the gypsum board surface in the detection image formed by unidirectional structured light, and accurately identify the type of defects.

[0060] The present invention uses a deep learning model to realize the automatic recognition of defects in the detection image formed by unidirectional structured light. On the basis of unidirectional structured light reducing the data processing volume and improving the defect detection efficiency, the defect detection efficiency is further improved.

[0061] The inspection images are preprocessed to uniform specifications before defect recognition.

[0062] The present invention uses a deep learning classifier model in unidirectional structured light defect detection to identify whether defects exist on the gypsum board surface in the detection image formed by unidirectional structured light, and accurately identify the type of defects. The details are as follows:

[0063] Use multiple classifiers to identify defects in the inspection image and obtain the defect detection results of the inspection image, including:

[0064] Input the inspection image into the multi-classifier, and the multi-classifier outputs the defect detection label of the inspection image;

[0065] Defect detection labels include gypsum board defect labels and gypsum board non-defect labels. Gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels.

[0066] The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively;

[0067] The gypsum board protrusion defect label corresponds to the test result of the gypsum board having a protrusion defect and the test result of the gypsum board having a concave defect.

[0068] Training of multiple classifiers, including:

[0069] Obtain a set of detection images as training samples;

[0070] Label a set of training samples to obtain the defect detection label of each training sample;

[0071] The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier;

[0072] The softmax classifier is used to classify the input items and output items to obtain a multi-classifier for defect recognition in the inspection image;

[0073] The multiple classifiers are:

[0074] Label = softmax(G);

[0075] Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

[0076] All training samples are normalized.

[0077] In actual use, when the gypsum board moves, the industrial camera collects the gypsum board image in real time, and then transmits the collected gypsum board defect detection image to the computer (data processor). The multi-classifier running in the computer identifies defects in the collected gypsum board defect detection image, and records the ID of the gypsum board corresponding to the detection image with the gypsum board protrusion defect label and the gypsum board concave defect label into the database. The database is queried. If a defect record is found in the current board, a rejection signal is sent to the rejection mechanism and an alarm is issued.

[0078] like Figure 2 As shown, the present invention provides a gypsum board defect detection system based on a unidirectional structured light gypsum board defect detection method, comprising:

[0079] a structured light emitter, configured to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly toward the gypsum board surface at a preset angle;

[0080] Industrial cameras, used to capture images formed by unidirectional structured light irradiating the surface of gypsum boards for detecting defects in gypsum boards;

[0081] A data processor is used to identify defects in the inspection image using multiple classifiers to obtain defect detection results of the inspection image;

[0082] The detection output portal is used to obtain the defect detection results of the gypsum board according to the defect detection results of the detection image.

[0083] In order to reduce detection costs and improve detection efficiency, the present invention forms unidirectional structured light on the surface of the gypsum board, uses a high-speed camera to shoot, and determines whether there are defects based on the degree of light bending of the structured light, thereby avoiding the need to arrange a large number of laser detection systems on the detection surface of the gypsum board, achieving cost reduction, reducing the amount of data processing, enhancing defect detection efficiency, and meeting the timeliness requirements of defect detection.

[0084] Furthermore, the present invention is combined with a classifier model in deep learning during unidirectional structured light defect detection to identify whether there are defects on the gypsum board surface in the detection image formed by unidirectional structured light, and accurately identify the type of defects.

[0085] The present invention uses a deep learning model to realize the automatic recognition of defects in the detection image formed by unidirectional structured light. On the basis of unidirectional structured light reducing the data processing volume and improving the defect detection efficiency, the defect detection efficiency is further improved.

[0086] The present invention forms unidirectional structured light on the surface of the gypsum board, uses a high-speed camera to capture it, and determines whether there are defects based on the degree of light bending of the structured light. This avoids the need to deploy a large number of laser detection systems on the inspection surface of the gypsum board, achieves cost reduction, reduces the amount of data processing, enhances defect detection efficiency, and meets the timeliness requirements of defect detection. The details are as follows:

[0087] The industrial camera is located below the surface of the gypsum board and vertically captures the detection image formed by the unidirectional structured light irradiated on the surface of the gypsum board.

[0088] The present invention uses a deep learning classifier model in unidirectional structured light defect detection to identify whether defects exist on the gypsum board surface in the detection image formed by unidirectional structured light, and accurately identify the type of defects. The details are as follows:

[0089] The data processor uses multiple classifiers to identify defects in the inspection image and obtains defect detection results for the inspection image, including:

[0090] Input the inspection image into the multi-classifier, and the multi-classifier outputs the defect detection label of the inspection image;

[0091] Defect detection labels include gypsum board defect labels and gypsum board non-defect labels. Gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels.

[0092] The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively;

[0093] The gypsum board protrusion defect label corresponds to the test result of the gypsum board having a protrusion defect and the test result of the gypsum board having a concave defect.

[0094] The data processor trains multiple classifiers, including:

[0095] Obtain a set of detection images as training samples;

[0096] Label a set of training samples to obtain the defect detection label of each training sample;

[0097] The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier;

[0098] The softmax classifier is used to classify the input items and output items to obtain a multi-classifier for defect recognition in the inspection image;

[0099] The multiple classifiers are:

[0100] Label = softmax(G);

[0101] Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

[0102] The data processor preprocesses the detection images into a unified specification and normalizes the training samples.

[0103] The preset angle is the optimal illumination angle of the structured light emitter that enables the defect detection network to achieve the best performance. The present invention constructs a layout model to determine the preset angle, that is, to determine the optimal illumination angle of the structured light emitter, specifically:

[0104] Randomly setting the illumination angle and the shooting angle as the learning angle, so that the structured light emitter and the high-speed industrial camera form structured light at the learning angle as the learning structured light, and the defect detection image as the learning image;

[0105] By using a pre-established defect detection model for image recognition of defect detection images, an angle evaluation is performed on the learning angle according to the learning structured light and the learning image to obtain an angle evaluation value of the learning angle;

[0106] Using a neural network, mapping learning is performed between learning angles and angle evaluation values ​​to obtain a layout model for optimal layout of illumination angles and shooting angles.

[0107] Randomly setting multiple illumination angles for the structured light emitter as first learning angles, and randomly setting multiple shooting angles for the high-speed industrial camera as second learning angles;

[0108] The structured light emitter is sequentially irradiated at a first learning angle to form structured light, which is used as a set of learning structured light, and the high-speed industrial camera is sequentially used to shoot a set of learning structured light at a second learning angle to obtain a set of defect detection images, which are used as a set of learning images.

[0109] Sequentially perform image recognition on the learning images formed by photographing the learning structured light formed at each first learning angle at each second learning angle using the defect detection model, and record the image recognition accuracy of the defect detection model for each learning structured light formed at the first learning angle on the learning images formed at each second learning angle as the image recognition accuracy of the learning structured light formed at each first learning angle;

[0110] sequentially taking the image recognition accuracy of the learning structured light formed at each first learning angle as the angle evaluation value of each first learning angle;

[0111] Sequentially photographing a learning image formed by the learning structured light formed by each first learning angle at each second learning angle, performing image recognition using the defect detection model, and recording the image recognition accuracy of the defect detection model for the learning image formed by the learning structured light formed by each first learning angle at each second learning angle as the image recognition accuracy of the learning image formed by each second learning angle;

[0112] The image recognition accuracy of the learning image formed at each second learning angle is sequentially used as the angle evaluation value of each second learning angle.

[0113] A BP neural network is used to construct a layout model for angle evaluation of illumination angle and shooting angle. The layout model is used to calculate the accuracy of the obtained structured light and defect detection images in the defect detection model according to arbitrary illumination angle and shooting angle, that is, the model performance of the defect detection model is calculated. The optimal layout is selected to achieve the best model performance of the defect detection model, thereby ensuring the gypsum board defect detection effect. The details are as follows:

[0114] Using a neural network, mapping learning is performed between learning angles and angle evaluation values ​​to obtain a layout model for optimal layout of illumination angles and shooting angles, including:

[0115] The first learning angle and the second learning angle are used as input items of the BP neural network, and the angle evaluation value of the first learning angle and the angle evaluation value of the second learning angle are used as output items of the BP neural network;

[0116] The BP neural network is used to map the input items of the BP neural network and the output items of the BP neural network to obtain a layout model;

[0117] The layout model is:

[0118] (P1,P2)=BP(D1,D2);

[0119] Where P1 is the angle evaluation value of the first learning angle, P2 is the angle evaluation value of the second learning angle, D1 is the first learning angle, D2 is the second learning angle, and BP is the BP neural network.

[0120] The present invention uses unidirectional structured light to form a detection image for detecting gypsum board defects, uses multiple classifiers to identify defects in the detection image, and obtains defect detection results of the detection image. The data processing volume is small, and the detection cost of the unidirectional structured light is low.

[0121] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. A gypsum board defect detection method based on unidirectional structured light, characterized in that: The following steps are involved: Using a structured light emitter to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly at a preset angle toward the gypsum board surface; Using an industrial camera to collect a detection image for detecting defects in the gypsum board formed by irradiating the unidirectional structured light on the surface of the gypsum board; Using multiple classifiers to perform defect recognition on the detection image to obtain a defect detection result of the detection image; According to the defect detection result of the detection image, a corresponding defect detection result of the gypsum board is obtained.

2. The gypsum board defect detection method based on unidirectional structured light according to claim 1, characterized in that: The inspection images are preprocessed to uniform specifications before defect recognition.

3. The gypsum board defect detection method based on unidirectional structured light according to claim 1, characterized in that: The method of using multiple classifiers to perform defect recognition on the detection image to obtain a defect detection result of the detection image includes: Inputting the detection image into a multi-classifier, and having the multi-classifier output a defect detection label of the detection image; The defect detection labels include gypsum board defect labels and gypsum board non-defect labels, and the gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels; The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively; The gypsum board protrusion defect label corresponds to the detection result of the gypsum board having a protrusion defect and the detection result of the gypsum board having a concave defect.

4. The gypsum board defect detection method based on unidirectional structured light according to claim 3, characterized in that: The training of the multiple classifiers includes: Obtain a set of detection images as training samples; Label a set of training samples to obtain the defect detection label of each training sample; The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier; Using a softmax classifier to perform classification learning on the input items and the output items to obtain a multi-classifier for defect recognition of the inspection image; The multi-classifier is: Label = softmax(G); Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

5. The gypsum board defect detection method based on unidirectional structured light according to claim 4, characterized in that: All training samples are normalized.

6. A gypsum board defect detection system based on unidirectional structured light, characterized in that: A gypsum board defect detection method based on unidirectional structured light according to any one of claims 1 to 5, wherein the gypsum board defect detection system comprises: a structured light emitter, configured to emit unidirectional structured light for detecting defects in the gypsum board from one side of the gypsum board conveying assembly toward the gypsum board surface at a preset angle; an industrial camera for collecting an inspection image formed by irradiating the gypsum board surface with the unidirectional structured light for detecting defects in the gypsum board; A data processor, configured to perform defect recognition on the detection image using a multi-classifier to obtain a defect detection result of the detection image; The detection output portal is used to obtain the defect detection result of the gypsum board according to the defect detection result of the detection image.

7. The gypsum board defect detection system based on unidirectional structured light according to claim 6, characterized in that: The industrial camera is located below the surface of the gypsum board and vertically captures a detection image formed by the unidirectional structured light irradiated on the surface of the gypsum board.

8. The gypsum board defect detection system based on unidirectional structured light according to claim 6, characterized in that: The data processor uses a multi-classifier to perform defect recognition on the detection image to obtain a defect detection result of the detection image, including: Inputting the detection image into a multi-classifier, and having the multi-classifier output a defect detection label of the detection image; The defect detection labels include gypsum board defect labels and gypsum board non-defect labels, and the gypsum board defect labels include gypsum board protrusion defect labels and gypsum board concave defect labels; The gypsum board defect label and the gypsum board non-defect label correspond to the test results of the gypsum board having defects and the test results of the gypsum board not having defects, respectively; The gypsum board protrusion defect label corresponds to the detection result of the gypsum board having a protrusion defect and the detection result of the gypsum board having a concave defect.

9. The gypsum board defect detection system based on unidirectional structured light according to claim 8, characterized in that: The data processor trains multiple classifiers, including: Obtain a set of detection images as training samples; Label a set of training samples to obtain the defect detection label of each training sample; The training samples are used as the input of the softmax classifier, and the defect detection labels of the training samples are used as the output of the softmax classifier; Using a softmax classifier to perform classification learning on the input items and the output items to obtain a multi-classifier for defect recognition of the inspection image; The multi-classifier is: Label = softmax(G); Where Label is the defect detection label, G is the detection image, and softmax is the softmax classifier.

10. The gypsum board defect detection system based on unidirectional structured light according to claim 6, characterized in that: The data processor preprocesses the detection image into a uniform specification and performs normalization processing on the training samples.

Images