A method for detecting wood damage defects based on point cloud data

Abstract

The present invention provides a method for detecting wood damage defects based on point cloud data, and the specific steps are as follows: step 1: data collection; step 2: preprocessing; step 3: effective data extraction; obtaining contour data of point cloud data; step 4: performing multiple mean filtering on each preprocessed point cloud contour data, and extracting residual values; step 5: defect determination; setting the threshold for detecting defects to C, and comparing the residual value with the set threshold; setting the threshold for detecting defects to C; when the residual value has a part greater than the set threshold C, it is determined that the corresponding point cloud contour has defects, otherwise it is determined to be a normal point cloud contour; the application of this technical solution can effectively improve the accuracy and production efficiency of damage defect detection in the forestry production process, and no longer needs to rely on manual marking of defects, which greatly saves human resource costs.

Description

Technical Field

[0001] The invention relates to the technical field of wood defect detection, and in particular to a wood damage defect detection method based on point cloud data. Background Art

[0002] In the wood furniture industry, the production of wood furniture must ensure the strength and structural integrity of the wood; the use of high-quality wood raw materials can effectively improve the production efficiency and quality of furniture. Therefore, accurately and quickly detecting damage and other defects on wood has become an important part of the industrial production process.

[0003] At present, many wood processing plants still use manual methods to mark defect locations during the production process. This method is not only inefficient and inaccurate, but also subjective. For example, stains on the wood can also have a certain impact on the judgment. Although machine vision-based detection methods have been used to detect wood defect areas, this method cannot measure the changes in the depth of the wood surface caused by damage defects. Therefore, developing a method for detecting wood damage defects is of great significance to the production process of the wood industry. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method for detecting wood damage defects based on point cloud data, and to determine whether there are damage defects based on the wood surface depth information provided by the three-dimensional point cloud data collected by the 3D laser sensor. First, each 3D point cloud contour data is subjected to multiple mean filtering in an iterative manner to construct a support contour close to the normal wood surface. The support contour data is then subtracted from the actual contour data to extract the residual value of the contour as the area of ​​interest for the wood damage defect. The residual value of the contour is compared with the set threshold to determine whether the contour contains defects. After all contour data of the 3D point cloud data are judged, the merging and filtering algorithm is used to calculate and obtain the final defect detection result.

[0005] To achieve the above purpose, the present invention adopts the following technical solution: a method for detecting wood damage defects based on point cloud data, the specific steps are as follows:

[0006] Step 1: Data collection; the wooden board is transported to the data collection area via a conveyor belt and triggers the laser sensor switch; the laser sensor then scans the upper and lower surfaces of the wood at a fixed sampling frequency and generates 3D point cloud data;

[0007] Step 2: Preprocessing: The collected H×W size point cloud data is used to filter out outliers in the wood point cloud data by using the quartile method to obtain valid wood point cloud data;

[0008] Step 3: Extract valid data; divide the H×N point cloud data according to the H dimension to obtain W i {i=0, 1, ..., H} is the contour data of the point cloud data;

[0009] Step 4: Perform multiple mean filtering on each preprocessed point cloud contour data using one-dimensional convolution calculation to construct the support contour and extract the residual value;

[0010] Step 5: Defect determination: The threshold for detecting defects is set to C, and the residual value R i =(r 1 , r 2 , ..., r N ) is compared with the set threshold to determine defects; the threshold for detecting defects is set to C; when the residual value is greater than the set threshold C, the corresponding point cloud contour is judged to have defects, otherwise it is judged to be a normal point cloud contour.

[0011] In a preferred embodiment, the residual value is calculated as follows:

[0012] Mean filtering: using Filter each contour to construct a support contour; where S i =s 1 ,s 2 , ..., s N is the filtering result, supporting contour data G i =(g 1 , g 2 , ..., g N ) is the initial point cloud contour data W i =(w 1 , w 2 , ..., w N ), {i=0,1,...,H}, and in the iterative process, update G i The value of

[0013] Set the size of the one-dimensional convolution kernel f(m) to K, W i is the initial point cloud contour data, N is the contour W i The number of point clouds;

[0014] Constructing support contours: Compare the filtered data S obtained by convolution i With point cloud support contour data G i ,pass

[0015]

[0016] The larger value of the two data is taken as the new support contour G′ i ;

[0017] The new support contour G′ is calculated i , G′ i Assign to G i , and repeat steps (5)-(7) T times to obtain the final timber support profile G i ;

[0018] The final timber support profile G i =(g 1 , g 2 , ..., g N ) and the initial point cloud contour data W i =(w 1 , w 2 , ..., w N ) is subtracted to obtain the residual value R i =(r 1 , r 2 , ..., r N )=(g 1 -w 1 , g 2 -w 1 , ..., g N -w N ).

[0019] In a preferred embodiment, specifically, the defect detection result is composed of a sequence of 0 and 1, denoted as A=(a 0 , a 1 , …, a h ), 0 represents a non-defective contour, and 1 represents a defective contour; in post-processing, the detected defect contours are merged using a merge operation. When the distance between two defect data contours is less than the set value M, they are merged into a defect area; after merging, when the defect area is less than the set value D, the defect area is filtered out as noise, thereby obtaining the final wood damage defect detection result.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1) The present invention effectively improves the accuracy and production efficiency of wood damage defect detection in the industrial processing and production process, and no longer needs to rely on manual marking of defects, which greatly saves human resource costs.

[0022] 2) Relying on obtaining wood three-dimensional point cloud data to analyze the depth changes on the wood surface and then detect wood damage defects can reduce the problem of false detection that is prone to occur in defect detection methods based on traditional machine vision.

[0023] 3) The point cloud contour data is detected one by one and the final result is corrected through merging and filtering operations to effectively ensure the accuracy of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flow chart of a support profile mean filtering algorithm according to a preferred embodiment of the present invention;

[0025] Figure 2 This is a flow chart of wood damage defect detection according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0026] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0029] A wood damage defect detection method based on point cloud data. The overall process is as follows: Figure 1 As shown, the specific steps are as follows:

[0030] Step 1: Data collection: The wood board is transported to the data collection area via a conveyor belt and triggers the laser sensor switch. The upper and lower laser sensors then scan the wood surface at a fixed sampling frequency and generate 3D point cloud data. The generated data is then transferred to a computer for subsequent processing.

[0031] Step 2: Preprocessing: For the collected point cloud data of 4000×450 size, outliers in the wood point cloud data are filtered out by the quartile method to obtain valid wood point cloud data.

[0032] Step 3: Extract valid data: Divide the point cloud data of 4000×450 size according to the dimension in the length direction to obtain W i {i=0, 1, ..., 4000} is the point cloud contour data.

[0033] Step 4: Use one-dimensional convolution calculation to perform multiple mean filtering on each preprocessed point cloud contour data to construct the support contour and extract the residual value. The algorithm flow is as follows: Figure 2 As shown, the specific steps are as follows:

[0034] 1) Mean filtering: using Filter each contour to construct a support contour. i =(s 1 ,s 2 , ..., s 450 ) is the filtering result, supporting contour data G i =(g 1 , g 2 , ..., g 450 ) is the initial W i =(w 1 , w 2 , ..., w 450 ), {i=1,...,4000}, and update G in the iteration process i The value of .

[0035] 2) Set the size of the one-dimensional convolution kernel f(m) to 60, W i is the initial point cloud contour data, 450 is the contour W i The number of point clouds.

[0036] 3) Construct support contour: Compare the filtered data S obtained by convolution i With point cloud support contour data G i ,pass

[0037]

[0038] The larger value of the two data is taken as the new support contour G′ i .

[0039] 4) Calculate the new support profile G′ i , G′ i Assign to G i , and repeat steps (5)-(7) 10 times to obtain the final timber support profile G i .

[0040] 5) Extract residual value: The final wood support profile G i =(g 1 , g 2 , ..., g 450 ) and the initial point cloud contour data W i =(w 1 , w 2 , ..., w 450) to obtain the residual value R i =(r 1 , r 2 , ..., r 450 )=(g 1 -w 1 , g 2 -w 1 , ..., g 450 -w 450 ).

[0041] Step 4: Defect determination: The threshold C for detecting defects is set to 1.2 mm, and the residual value R i =(r 1 , r 2 , ..., r 450 ) is compared with the set threshold to determine defects. When the residual value is greater than the set threshold, the corresponding point cloud contour is determined to have defects, otherwise it is judged to be a normal point cloud contour.

[0042] Step 5: The defect detection result is composed of a sequence of 0 and 1, denoted as A = (a 0 , a 1 , …, a 4000 ), 0 represents a defect-free contour and 1 represents a defective contour.

[0043] Step 6: In post-processing, the detected defect contours are merged using the merge operation. When the distance between two defect data contours is less than the set value of 30 mm, they are merged into one defect area.

[0044] Step 7: After merging, when the defective area is smaller than the set value of 5mm, the defective area is filtered out as noise to obtain the final wood damage defect detection result.

Claims

1. A method for detecting wood damage defects based on point cloud data. It is characterized in that The specific steps are as follows: Step 1: Data collection; the wooden board is transported to the data collection area via a conveyor belt and triggers the laser sensor switch; the laser sensor then scans the upper and lower surfaces of the wood at a fixed sampling frequency and generates 3D point cloud data; Step 2: Preprocessing: The collected H×W size point cloud data is used to filter out outliers in the wood point cloud data by using the quartile method to obtain valid wood point cloud data; Step 3: Extract valid data; divide the H×N point cloud data according to the H dimension to obtain W i {i=0,1,…,H} is the contour data of point cloud data; Step 4: Perform multiple mean filtering on each preprocessed point cloud contour data using one-dimensional convolution calculation to construct the support contour and extract the residual value; Step 5: Defect determination; The threshold for detecting defects is set to C, and the residual value R i =(r 1 ,r 2 ,…,r N ) Defects are determined by comparing with the set threshold; the threshold for detecting defects is set to C; when the residual value is greater than the set threshold C, the corresponding point cloud contour is judged to have defects, otherwise it is judged to be a normal point cloud contour; The calculation steps of the residual value are as follows: Mean filtering: using Filter each contour to construct a support contour; where S i =s 1 ,s 2 ,…,s N is the filtering result, supporting contour data G i =(g 1 ,g 2 ,…,g N ) is the initial point cloud contour data W i =(w 1 ,w 2 ,…,w N ),{i=0,1,…,H}, and update G in the iterative process i The value of Set the size of the one-dimensional convolution kernel f(m) to K, and N to the contour W i The number of point clouds; Constructing support contours: Compare the filtered data S obtained by convolution i With point cloud support contour data G i ,pass The larger value of the two data is taken as the new support contour G' i ; The new support profile G' will be calculated i , G' i Assign to G i , and repeat steps (5)-(7) T times to obtain the final timber support profile G i ; The final timber support profile G i =(g 1 ,g 2 ,…,g N ) and the initial point cloud contour data W i =(w 1 ,w 2 ,…,w N ) is subtracted to obtain the residual value R i =(r 1 ,r 2 ,…,r N )=(g 1 -w 1 ,g 2 -w 1 ,…,g N -w N ).

2. According to the method for detecting wood damage defects based on point cloud data in claim 1, It is characterized in that Specifically, the defect detection result is composed of a sequence of 0 and 1, denoted as A=(a 0 ,a 1 ,…,a h ), 0 represents a non-defective contour, and 1 represents a defective contour; in post-processing, the detected defect contours are merged using a merge operation. When the distance between two defect data contours is less than the set value M, they are merged into a defect area; after merging, when the defect area is less than the set value D, the defect area is filtered out as noise, thereby obtaining the final wood damage defect detection result.

Images