A quality inspection, grading and intelligent sorting system for peeled plates

Abstract

This invention discloses a quality inspection, grading, and intelligent sorting system for veneer-cut plates, relating to the field of image recognition technology. It solves the problem of difficult detection and classification of veneer-cut plates due to their large area and the variety and complexity of surface defects. The invention includes a data acquisition unit, an image processing unit, and a classification and alarm unit. The data acquisition unit acquires images of the veneer-cut plates and stores them in a storage module or transmits them wirelessly to the image processing unit for further processing. The image processing unit uses a defect recognition model trained in the image processing unit to identify defects in the veneer-cut plates. After inputting images of the veneer-cut plates, the system outputs the types and quantities of defects, as well as image cropping of each identified defect. The similarity of the cropped images within the same defect type is then calculated, and a reliability evaluation coefficient is derived by combining the confidence level. Finally, a classification standard is constructed based on the model output and the reliability evaluation coefficient to classify the veneer-cut plates.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a quality detection, grading and intelligent sorting system for rotary cut plates. Background Art

[0002] Before being put into use, peeled boards must undergo a rigorous quality inspection and grading process to ensure that they meet specific industry standards and usage requirements. This process traditionally relies on manual visual inspection, where workers need to carefully observe each peeled board to identify and count defects such as knots and shadows on the board. This work is not only labor-intensive, but also requires extremely high vision and concentration from workers. Workers are prone to visual fatigue after performing this highly repetitive and high-precision work for a long time, which not only affects the accuracy of the inspection, but also poses a potential risk to workers' health. In the current manual inspection process, it takes about 12 to 15 seconds to inspect each peeled board, which means that a worker can inspect about 5 boards in one minute. Even experienced and skilled workers can only increase the inspection speed to 8 to 10 boards per minute, but this speed still cannot meet the needs of large-scale production, and workers cannot maintain this high-intensity work state continuously. In addition, another limitation of manual inspection is that workers can only count wood knots and other defects, but cannot provide more in-depth quantitative analysis, which makes the inspection results often lack accuracy and difficult to meet the industry's detailed standards for peeled board quality inspection.

[0003] Due to the subjectivity and fatigue of manual inspection, grading errors often occur, which not only affects the quality of the peeled board, but may also have a negative impact on the processing and use of subsequent products. Moreover, the peeled board products are objectively present in a large area. The large product area means that there are many types of surface defects, resulting in diverse inspection objects, various types of surface defects, various forms, and complex backgrounds. Moreover, since defects are randomly generated and randomly distributed, it is difficult to find a "standard" image as a reference, which makes it difficult to detect and classify defects.

[0004] In view of this, a peeled plate quality inspection, grading and intelligent sorting system is needed. Summary of the invention

[0005] In view of the problem in the prior art that it is difficult to detect and classify peeled boards due to the large area of ​​peeled boards and the many and complex types of surface defects, the present invention provides a peeled board quality detection grading and intelligent sorting system, which can identify the types and quantities of defects, as well as confidence and defect location image capture based on a trained defect recognition model, and then calculate the similarity of each captured image of the same defect type, and derive a reliability evaluation coefficient based on the confidence, and finally construct a classification standard based on the model output results and the reliability evaluation coefficient to classify the peeled boards. The specific technical solution is as follows:

[0006] A peeled plate quality inspection, grading and intelligent sorting system, comprising:

[0007] A data acquisition unit, used to acquire images of the peeled plate and store them in a storage module or transmit them to an image processor via a wireless transmission module for further processing;

[0008] The image processing unit includes a model training and use module, in which a defect recognition model is trained for defect recognition of the rotary cut plate. The input of the defect recognition model is a rotary cut plate image, and the output is the types and quantities of defects of the rotary cut plate, as well as image captures of each identified defect.

[0009] The classification and alarm unit is connected to the image processing unit, receives the defect recognition results output from the image processing unit and performs classification and alarm.

[0010] Preferably, the image processing unit further includes a similarity acquisition module, which is used to collect pictures of all defects of the same type after the defect recognition model is passed, and calculate the similarity between the pictures to obtain a comprehensive similarity coefficient, which is as follows:

[0011] First, calculate the similarity between the two images. The similarity between image a and image b is calculated as follows:

[0012]

[0013] X=[x1,x2,…,x i ,…,x n ]

[0014] Y=[y1,y2,…,y i ,…,y n ]

[0015] In the formula, x i ∈X,y i ∈Y, X and Y are the texture feature vectors of the two images used for comparison;

[0016] Then, the similarity comprehensive coefficient is calculated, and the calculation formula is as follows:

[0017]

[0018] Where M is the number of defects of defect type k identified, that is, the number of images of defect k obtained, a is the first a defect images, b is the bth defect image, cosθ a,b is the similarity between the a-th picture and the b-th picture.

[0019] Preferably, the image processing unit further includes a reliability statistics module for obtaining confidence and similarity, and then obtaining a reliability evaluation coefficient reflecting the reliability of the output result which is finally output to the terminal. The calculation of the reliability evaluation coefficient is specifically as follows:

[0020]

[0021] In the formula, Q is the reliability evaluation coefficient, is the confidence comprehensive coefficient of the k-th defect type, is the comprehensive similarity coefficient of the k-th defect type, μ1 is the comprehensive confidence coefficient of the k-th defect type , μ2 is the weight coefficient of the comprehensive similarity coefficient of the kth defect type;

[0022] Among them, the confidence comprehensive coefficient The calculation formula is specifically expressed as follows:

[0023]

[0024] Where M is the number of defects of defect type k identified, ρ c Confidence of the cth defect type k identified.

[0025] Preferably, the construction process of the defect recognition model is as follows:

[0026] S1: Data preparation and preprocessing: Collect images of peeled plates with various defects and ensure data diversity and representativeness. Resize, normalize, and convert the images to meet model input requirements;

[0027] S2: Data annotation: Accurately annotate the collected images, including the category and location of the defects, marked in the form of rectangular boxes;

[0028] S3: Data enhancement: Use data enhancement techniques to increase data diversity, including rotation, flipping, and deformation;

[0029] S4: Model selection and training: Select deep learning models for multi-class classification and object detection, and train the models using labeled data, including category classification and bounding box prediction;

[0030] S5: Post-processing of results: Post-process the output of the model to extract categories, bounding boxes, and confidence scores, and convert category IDs into corresponding Chinese names;

[0031] Preferably, the defects include blunt ballast, wood brushes, wood knots, cracks, decay, mildew, worm holes, scars, gaps, cutting marks, and feeder marks.

[0032] Preferably, the classification and alarm unit includes a classification module, and the classification module performs classification according to the following classification principles:

[0033]

[0034] In the formula, L is the final output classification, L P is the Pth classification type set, λ P is the threshold value, δ k is the influence coefficient of the kth defect type, U is the number of defect types, Q max It is the maximum reliability evaluation coefficient, that is, the value of the reliability evaluation coefficient when both the similarity and the confidence are at their maximum values.

[0035] Preferably, the classification and alarm unit includes an alarm module, in which alarm classification and non-alarm classification are set. When the classification result output by the classification module is an alarm classification, an alarm operation is performed, otherwise, no alarm operation is performed.

[0036] Preferably, the alarm operation includes using sound and light alarms and sending an alarm message to the terminal.

[0037] A computer-readable storage medium includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned peeled plate quality detection, grading and intelligent sorting system.

[0038] A processor is used to run a program, wherein when the program is run, the peeled plate quality detection, grading and intelligent sorting system as described above is executed.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention is provided with a data acquisition unit, an image processing unit and a classification and alarm unit. The data acquisition unit collects pictures of the peeled plate and stores them in a storage module or transmits them to the image processing unit through a wireless transmission module for further processing. The defect recognition model trained in the image output unit is used to identify the defects of the peeled plate. After the peeled plate picture is input, the types and quantities of defects of the peeled plate and the picture captures of each identified defect are output, and finally classification and alarm are performed based on this.

[0041] 2. The present invention can identify the types and quantities of defects, as well as confidence levels and defect location image capture based on a trained defect recognition model, and then calculate the similarity of each captured image of the same defect type, and derive a reliability evaluation coefficient based on the confidence level. Finally, a classification standard is constructed based on the model output results and the reliability evaluation coefficient to classify the rotary cut plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0043] Figure 1 It is the system principle diagram of the present invention;

[0044] Figure 2 The figure is a flow chart for constructing the defect recognition model of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0047] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0048] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0049] In one embodiment of the present invention, a system for quality inspection, grading and intelligent sorting of peeled plates is provided. Figure 1 As shown, including:

[0050] The data acquisition unit includes a camera, a storage module and a wireless transmission module. The camera is used to collect images of the peeling plate and store them in the storage module or transmit them to the image processing unit through the wireless transmission module for further processing.

[0051] The image processing unit includes a model training and use module, a similarity acquisition module and a reliability statistics module. The model training and use module trains a defect recognition model for identifying defects in peeled boards, and the input of the defect recognition model is a peeled board image, and the output is the type and number of peeled board defects, as well as image captures of each identified defect.

[0052] like Figure 2 As shown, the construction process of the defect recognition model is as follows:

[0053] S1: Data preparation and preprocessing: Collect images of peeled boards with various defects and ensure data diversity and representativeness. Resize, normalize, and convert the images to fit the model input requirements. Defects include blunt slag, wood brush, wood knots, cracks, rot, mold, wormholes, scars, gaps, cutting marks, and feeder marks.

[0054] S2: Data annotation: Accurately annotate the collected images, including the category and location of the defects, marked in the form of rectangular boxes.

[0055] S3: Data enhancement: Use data enhancement techniques such as rotation, flipping, deformation, etc. to increase data diversity and improve model generalization ability.

[0056] S4: Model selection and training: Select a deep learning model suitable for multi-category classification and target detection, such as the YOLO series model and the Cascaded Twin Dense Network (SCSEG-Net). Use labeled data to train the model, including category classification and bounding box prediction.

[0057] S5: Model optimization: Introduce attention mechanisms, such as convolutional block attention module (CBAM), to enhance the model's ability to extract defect features. For similar defects, a cascade network structure can be used to first perform a rough segmentation of the defect area, and then perform fine classification on the segmented defect image.

[0058] S6: Model evaluation and adjustment: Use indicators such as precision, recall, and F1 score to evaluate model performance. Adjust model parameters and structure based on the evaluation results to optimize model performance.

[0059] S7: Post-processing of results: Post-process the output of the model to extract categories, bounding boxes, and confidence scores, and convert category IDs to corresponding Chinese names for easy understanding.

[0060] Through the above steps, a model can be built that can effectively identify multiple defects in peeled plates and output the defect type, quantity, and location screenshots of each defect. The key lies in meticulous data preprocessing and annotation, data enhancement, and continuous optimization and updating of the model. Through these methods, the model's ability to distinguish similar defects can be improved and the accuracy of identification can be guaranteed.

[0061] Through the defect recognition model, we can get the defect type, the location screenshot of the defect, the confidence score of the defect recognition, and then we can get the number of a certain defect type. The collected peeled plate images are input into the defect recognition model to obtain the above data, which is convenient for further analysis and alarm.

[0062] The similarity acquisition module is used to collect image captures of all defects of the same type after the defect recognition model, further analyze the similarity between these images, and further verify the recognition results of the defect recognition model. In other words, in addition to the confidence indicator of the defect recognition model output, each output result identified as the same defect type is further verified to check the similarity of images of the same defect type, and then further judge whether these defects are similar, and then further judge the accuracy of the output results of the defect recognition model.

[0063] The similarity acquisition module is based on the cosine similarity between the texture features of the two images. After obtaining the similarity between the two images, the average value of these similarities is calculated, and then a similarity comprehensive coefficient is synthesized. The similarity comprehensive coefficient is used to reflect the similarity between the images of this type, and then reflect the accuracy or credibility of the recognition results of the defect recognition model.

[0064] The similarity between image a and image b is calculated as follows:

[0065]

[0066] In the formula, x i ∈X,y i ∈Y, X and Y are the texture feature vectors of the two images used for comparison, which can be expressed as: X = [x1, x2, ..., x i ,…,x n ],Y=[y1,y2,…,y i ,…,y n ].

[0067] Furthermore, the comprehensive similarity coefficient of defect k is the average value of the similarities calculated after the above two combinations. The specific calculation formula is as follows:

[0068]

[0069] Where M is the number of defects of defect type k identified, that is, the number of images of defect k obtained, a is the first a defect images, b is the bth defect image, cosθ a,b is the similarity between the a-th picture and the b-th picture.

[0070] The reliability statistics module is used to obtain the confidence and similarity, and then obtain the reliability evaluation coefficient that reflects the reliability of the output result that is finally output to the terminal. The calculation of the reliability evaluation coefficient is as follows:

[0071]

[0072] In the formula, Q is the reliability evaluation coefficient, is the confidence comprehensive coefficient of the k-th defect type, is the comprehensive similarity coefficient of the k-th defect type, μ1 is the comprehensive confidence coefficient of the k-th defect type , μ2 is the weight coefficient of the comprehensive similarity coefficient of the kth defect type.

[0073] Among them, the confidence comprehensive coefficient The calculation formula is specifically expressed as follows:

[0074]

[0075] Where M is the number of defects of defect type k identified, ρ c Confidence of the cth defect type k identified.

[0076] The classification and alarm unit includes a classification module and an alarm module, wherein the classification module is connected to the image processing unit, receives the defect recognition result and the reliability of the defect recognition result output from the image processing unit, and performs classification and alarm based on this. Specifically, the classification module receives the defect type k and its quantity M output from the defect recognition model, and receives the reliability evaluation coefficient output from the reliability statistics module. And the classification and alarm operation is performed based on the defect type k and its quantity M, and the reliability evaluation coefficient Q.

[0077] Among them, the classification is carried out according to the following classification principles:

[0078]

[0079] In the formula, L is the final output classification, L P is the Pth classification type set, λ P is the threshold value, δ k is the influence coefficient of the kth defect type, U is the number of defect types, Q maxIt is the maximum reliability evaluation coefficient, that is, the value of the reliability evaluation coefficient when both the similarity and the confidence are at their maximum values.

[0080] The alarm module is set with alarm classification and non-alarm classification, that is, the specific classifications L1-L p In the example, several Lx (x = 1, 2, ..., p) are set as alarm categories according to the actual situation. When the final output classification result is an alarm category, an alarm operation is performed, otherwise no alarm operation is performed. The alarm operation includes using sound and light alarms and sending alarm messages to the terminal.

[0081] The terminal is connected to the data acquisition unit, the image processing unit and the classification and alarm unit, and is used to receive process data and result data connected to the data acquisition unit, the image processing unit and the classification and alarm unit. The process data includes the collected original data, and the result data includes the output data of each unit.

[0082] The present invention is provided with a data acquisition unit, an image processing unit and a classification and alarm unit. The data acquisition unit collects pictures of rotary cut plates and stores them in a storage module or transmits them to an image processing unit through a wireless transmission module for further processing. The defect recognition model trained in the image output unit is used to recognize defects of the rotary cut plates. After the picture of the rotary cut plate is input, the types and quantities of defects of the rotary cut plate and the picture capture of each identified defect are output. Then, the similarity of each captured picture of the same defect type is calculated, and the reliability evaluation coefficient is obtained in combination with the confidence level. Finally, a classification standard is constructed based on the model output result and the reliability evaluation coefficient to classify the rotary cut plates.

[0083] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0084] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.

[0085] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0086] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A quality inspection, grading and intelligent sorting system for peeled plates, characterized in that: include: A data acquisition unit, used to acquire images of the peeled plate and store them in a storage module or transmit them to an image processor via a wireless transmission module for further processing; An image processing unit, comprising a model training and use module, wherein a defect recognition model is trained in the model training and use module for performing defect recognition of the peeled plate, and the input of the defect recognition model is a peeled plate image, and the output is the types and quantities of defects of the peeled plate, as well as image captures of each identified defect; The classification and alarm unit is connected to the image processing unit, receives the defect recognition results output from the image processing unit and performs classification and alarm.

2. The peeled board quality inspection, grading and intelligent sorting system according to claim 1 is characterized in that: The image processing unit also includes a similarity acquisition module, which is used to collect pictures of all defects of the same type after the defect recognition model is passed, and calculate the similarity between the pictures to obtain a similarity comprehensive coefficient, which is as follows: First, calculate the similarity between the two images. The similarity between image a and image b is calculated as follows: X=[x1,x2,…,x i ,…,x n ] Y=[y1,y2,…,y i ,…,y n ] In the formula, x i ∈X,y i ∈Y, X and Y are the texture feature vectors of the two images used for comparison; Then, the similarity comprehensive coefficient is calculated, and the calculation formula is as follows: Where M is the number of defects of defect type k identified, that is, the number of images of defect k obtained, a is the first a defect images, b is the bth defect image, cosθ a,b is the similarity between the a-th picture and the b-th picture.

3. The peeled board quality inspection, grading and intelligent sorting system according to claim 1 is characterized in that: The image processing unit further includes a reliability statistics module for obtaining confidence and similarity, and then obtaining a reliability evaluation coefficient reflecting the reliability of the output result which is finally output to the terminal. The calculation of the reliability evaluation coefficient is specifically as follows: In the formula, Q is the reliability evaluation coefficient, is the confidence comprehensive coefficient of the k-th defect type, is the comprehensive similarity coefficient of the k-th defect type, μ1 is the comprehensive confidence coefficient of the k-th defect type , μ2 is the weight coefficient of the comprehensive similarity coefficient of the kth defect type; Among them, the confidence comprehensive coefficient The calculation formula is specifically expressed as follows: Where M is the number of defects of defect type k identified, ρ c Confidence of the cth defect type k identified.

4. A peeled board quality inspection, grading and intelligent sorting system according to any one of claims 1 to 3, characterized in that: The construction process of the defect recognition model is as follows: S1: Data preparation and preprocessing: Collect images of peeled plates with various defects and ensure data diversity and representativeness. Resize, normalize, and convert the images to meet model input requirements; S2: Data annotation: Accurately annotate the collected images, including the category and location of the defects, marked in the form of rectangular boxes; S3: Data enhancement: Use data enhancement techniques to increase data diversity, including rotation, flipping, and deformation; S4: Model selection and training: Select deep learning models for multi-class classification and object detection, and train the models using labeled data, including category classification and bounding box prediction; S5: Post-processing of results: Post-process the output of the model to extract categories, bounding boxes, and confidence scores, and convert category IDs to corresponding Chinese names.

5. The peeled board quality inspection, grading and intelligent sorting system according to claim 4 is characterized in that: The defects include dull ballast, wood brushes, wood knots, cracks, rot, mildew, worm holes, scars, gaps, cutting marks, and feeder marks.

6. The peeled board quality inspection, grading and intelligent sorting system according to claim 3 is characterized in that: The classification and alarm unit includes a classification module, which classifies according to the following classification principles: In the formula, L is the final output classification, L P is the Pth classification type set, λ P is the threshold value set, δ k is the influence coefficient of the kth defect type, U is the number of defect types, Q max It is the maximum reliability evaluation coefficient, that is, the value of the reliability evaluation coefficient when both the similarity and the confidence are at their maximum values.

7. The peeled board quality inspection, grading and intelligent sorting system according to claim 6 is characterized in that: The classification and alarm unit includes an alarm module, in which alarm classification and non-alarm classification are set. When the classification result output by the classification module is an alarm classification, an alarm operation is performed, otherwise, no alarm operation is performed.

8. The peeled board quality inspection, grading and intelligent sorting system according to claim 7 is characterized in that: The alarm operation includes using sound and light alarms and sending alarm messages to the terminal.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the peeled plate quality detection, grading and intelligent sorting system described in claim 1.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the peeled board quality detection, grading and intelligent sorting system described in claim 1.

Images