A multi-piece wood abnormal working condition detection method based on three-dimensional data

Abstract

The present application relates to a kind of based on three-dimensional data multi-piece wood abnormal condition detection method.First, the three-dimensional data collected is filtered and denoised using quartile method, removes outlier on background, then extracts the effective data of each contour, and then the number of each contour effective data and the size of the three-dimensional data collected are compared and analyzed to set the threshold condition for judging abnormality, and finally determine the current condition.The present application is applied to wood anomaly detection, which can effectively remove the collected multi-piece wood abnormal data, provide convenience for subsequent detection operations based on wood three-dimensional data, and improve industrial production efficiency.

Description

Technical Field

[0001] This invention relates to the field of timber anomaly detection technology, and in particular to a method for detecting abnormal working conditions of multiple timber pieces based on three-dimensional data. Background Technology

[0002] During the wood processing process, the processing system can only process a single piece of wood. To ensure the normal operation of the wood processing system, it is necessary to analyze the collected three-dimensional data and detect abnormal working conditions of multiple pieces of wood.

[0003] In the process of acquiring 3D data of timber in industrial settings, manual verification of the acquisition process is required. Due to the high speed of production lines and the difficulty for workers to maintain focus for extended periods, multiple pieces of timber may pass through the acquisition system simultaneously, resulting in a single 3D data set containing anomalies from multiple timbers. Industrial sites require high-speed, high-precision multi-timber anomaly detection technology to replace manual inspection, improve production efficiency, and avoid multiple timber anomalies caused by human error.

[0004] The filtering algorithm improves the identification of the start and end points of each row of valid data in the multi-piece timber anomaly detection algorithm, avoiding interference from outliers. Secondary analysis and judgment of abnormal conditions ensure the accuracy of multi-piece timber anomaly detection. This multi-piece timber anomaly detection method plays a crucial role in timber data acquisition and processing.

[0005] Chinese patent application number CN201610288409.7, entitled "A Timber Counting Method Based on Contour Optimization," describes a method for counting timber. This method involves resizing a timber image, extracting the timber target image from the transformed image using a color model, extracting edge information within the image using an edge operator, fusing the extracted timber target image and edge information to obtain a pre-segmented image, and extracting the contours of candidate targets. The extracted outer and inner contours are then optimized using erosion and dilation, respectively. This process is repeated until the number of extracted outer contours remains constant and no inner contours are found. However, this method may be affected by the background color when extracting the timber target image.

[0006] Chinese patent application number CN201210455157.4, entitled "Steel Counting System and Method," describes a counting system comprising an image acquisition device, a host computer module, and a power supply. The image acquisition device is connected to the host computer module to acquire real-time images of steel on the production line and send them to the host computer module. The host computer module analyzes the acquired images to obtain steel count statistics. The power supply is connected to both the image acquisition device and the host computer module to provide power to them. This invention achieves automatic counting of the quantity, location, and dimensions of steel; however, the counting method is complex, time-consuming, inefficient, and detrimental to industrial production processes. Summary of the Invention

[0007] The purpose of this invention is to provide a method for detecting abnormal working conditions of multiple timber pieces based on three-dimensional data. First, the acquired three-dimensional data is filtered and denoised using the quartile method to remove outliers from the background. Then, the valid data for each contour is extracted. Next, by comparing the number of valid data points for each contour with the size of the acquired three-dimensional data, a threshold condition for judging anomalies is set, and finally, the current working condition is determined. Applying this invention to timber anomaly detection can effectively remove abnormal data from multiple timber pieces, facilitating subsequent detection operations based on three-dimensional timber data and improving industrial production efficiency.

[0008] To achieve the above objectives, the technical solution of this invention is: a method for detecting abnormal working conditions of multiple timber pieces based on three-dimensional data. The method, applied to timber anomaly detection, firstly uses the quartile method to filter and denoise the acquired three-dimensional data to remove outliers in the background. Then, it extracts the effective data for each contour. Using the number of effective data points for each contour and the size of the acquired three-dimensional data, a threshold condition for judging anomalies is set through comparative analysis, and finally, the current working condition is determined. The specific implementation steps of this method are as follows:

[0009] (1) Data acquisition: The wooden board is conveyed to the data collection area by a conveyor belt and the depth camera is activated; then the depth camera scans the upper and lower surfaces of the wood at a fixed sampling frequency and generates three-dimensional data.

[0010] (2) Preprocessing: The collected H×W three-dimensional data are filtered according to the quartile method to remove abnormal data.

[0011] (3) Effective data extraction: The three-dimensional data of size H×W is divided according to the H dimension to obtain W. i {i = 0, 1, ..., H} represents the contour data of the three-dimensional data;

[0012] (4) Using the formula Record the valid data positions for each contour, where the starting position is the minimum valid data position L for each contour. i The termination position is the maximum effective data position R for each contour. i And calculate the effective data width S for each contour. i =R i -L i ,{i=0,1,...,H};

[0013] (5) Through the formula Calculate the total number of valid values ​​P between the start and end positions of the profile. i ,{i=0,1,...,H};where x ij {i = 0, 1, ..., H, j = 0, 1, ..., W} represents the value in the i-th row and j-th column of the three-dimensional data;

[0014] (6) Set the initial count value n = 0, when the total number of valid data P i {i = 0, 1, ..., H} is less than M*S i When {i=0,1,...,H}, the count value n=n+1; where M is the set width scaling factor;

[0015] (7) Compare the count value n with the H value of the three-dimensional data of the wood. If the value of n is greater than N*H, it is determined that there are multiple abnormal data of wood pieces; where N is the set abnormal contour ratio coefficient.

[0016] (8) The following are the specific steps for the secondary judgment of the three-dimensional data of wood that are detected as normal:

[0017] (8.1) Remove the first and last D rows of contour data from the three-dimensional wood data to obtain new three-dimensional contour data W. i ,{i=D,D+1,…,HD};

[0018] (8.2) Calculate the total number of valid values ​​P for each contour data. i The quadratic gradient value Y of {i=D,D+1,...,HD} i ={[(P i+1 -P i ) / 2]-[(P i+3 -P i+2 ) / 2] / 2},{i=D,D+1,...,HD};

[0019] (8.3) Set the initial count value m = 0 and the gradient threshold K. When the second gradient value Y i When the count value of {i = D, D+1, ..., HD} is greater than the set threshold K, the count value is m = m+1.

[0020] (8.4) Compare the cumulative value m with the set threshold T. If m>T, it is determined that there are multiple abnormal data of wood pieces. Otherwise, the three-dimensional data of wood is normal data.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1) This invention enables the detection of multiple pieces of wood, effectively identifying abnormal working conditions during the processing, and effectively ensuring the stability and accuracy of wood processing.

[0023] 2) Relying on the collected 3D data, the contour-by-contour detection method is adopted, which has a fast detection speed. This not only ensures the accuracy of anomaly detection for multiple pieces of wood, but also meets the real-time requirements in industrial scenarios.

[0024] 3) By setting judgment thresholds and using a dual detection strategy, it is possible to detect anomalies in multiple pieces of wood of different sizes, effectively ensuring the subsequent handling of abnormal working conditions. Attached Figure Description

[0025] Figure 1 This is a flowchart of the abnormal working condition detection process for multiple pieces of wood according to the present invention.

[0026] Figure 2 This is a flowchart of the secondary testing process for multiple pieces of wood according to the present invention. Detailed Implementation

[0027] The above and other advantages of the present invention will become more apparent from the following detailed description of preferred embodiments in conjunction with the accompanying drawings. The present invention provides a method for detecting abnormal working conditions of multiple timber blocks based on three-dimensional data. The overall process is as follows: Figure 1 As shown, the specific steps are as follows:

[0028] (1) Data acquisition: The wooden board is conveyed to the data collection area via a conveyor belt, which triggers the switch of the depth camera. The depth camera then scans the upper and lower surfaces of the wood at a fixed sampling frequency and generates three-dimensional data.

[0029] (2) Preprocessing: The collected 3D data of size 3500×400 is filtered according to the quartile method to remove abnormal data.

[0030] (3) Effective data extraction: The 3D data of size 3500×400 is divided into H dimensions to obtain W. i {i=0,1,...,3500} represents the contour data of the three-dimensional data.

[0031] (4) Using the formula Record the valid data positions for each contour, where the starting position is the minimum valid data position L for each contour. iThe termination position is the maximum effective data position R for each contour. i And calculate the effective data width S for each contour. i =R i -L i ,{i=0,1,...,3500}.

[0032] (5) Through the formula Calculate the total number of valid values ​​P between the start and end positions of the profile. i {i = 0, 1, ..., 3500}. Where x ij {i = 0, 1, ..., 3500, j = 0, 1, ..., 400} represents the value in the i-th row and j-th column of the three-dimensional data.

[0033] (6) Set the initial count value n = 0, when the total number of valid data P i {i=0,1,...,3500} is less than M*S i When {i=0,1,...,3500}, the count value n=n+1. The width scaling factor M is set to 0.99.

[0034] (7) Compare the count value n with the length dimension 3500 of the three-dimensional data of the wood. If the value of n is greater than N*3500=1400, it is judged as multiple abnormal wood data. The abnormal contour ratio coefficient N is set to 0.4.

[0035] (8) A secondary judgment is made on the three-dimensional data of the wood that is detected as normal in the above steps, as shown in the flowchart. Figure 2 As shown. The steps are as follows:

[0036] (9) Remove the first and last 20 rows of contour data from the wood 3D data to obtain the new 3D contour W. i ,{i=20,21,…,3480}.

[0037] (10) Calculate the total number of valid values ​​P for each contour data. i The gradient value Y of {i = 20, 21, ..., 3480} i ={[(P i+1 -P i ) / 2]-[(P i+3 -P i+2 ) / 2] / 2},{i=20,21,...,3480}.

[0038] (11) Set the initial count value m = 0 and the gradient threshold K = 30. When the second gradient value Y i When {i = 20, 21, ..., 3480} is greater than the threshold 30, the count value m = m + 1.

[0039] (12) Compare the cumulative value m with the set threshold T = 2. If m > 2, it is determined that there are multiple abnormal data of wood pieces. Otherwise, the three-dimensional data of the wood is normal data.

[0040] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for detecting abnormal working conditions of multi-piece wood based on three-dimensional data, characterized in that, Applied to timber anomaly detection, the method first uses the quartile method to filter and denoise the acquired 3D data to remove outliers in the background. Then, it extracts the valid data for each contour. By comparing the number of valid data points for each contour with the size of the acquired 3D data, a threshold condition for anomaly detection is set, and finally, the current working condition is determined. The specific implementation steps of this method are as follows: (1) Data acquisition: The wooden board is conveyed to the data collection area by a conveyor belt and the depth camera is activated; then the depth camera scans the upper and lower surfaces of the wood at a fixed sampling frequency and generates three-dimensional data. (2) Preprocessing: The collected H×W three-dimensional data are filtered according to the quartile method to remove abnormal data. (3) Effective data extraction: divide the three-dimensional data of size HxW according to H dimensions, and obtain W i {i=0, 1,..., H} is the contour data of the three-dimensional data; (4) using the formula record the valid data position of each profile, where the start position is the minimum value L of the valid data position of each profile i , the end position is the maximum value R of the valid data position of each profile i , and the valid data width S of each profile is calculated i = R i - L i , i = 0, 1,..., H; (5) The total number of effective values P between the start and end positions of the contour is calculated by the formula P = ∑ x (i), i = 0, 1,..., H i where x (i) is the value of the i-th pixel in the contour ij where x (i, j) is the value of the i-th row and j-th column in the three-dimensional data (6) set initial count value n=0, when total number of valid data P i , i=0, 1,..., H} is less than M*S i , i=0, 1,..., H}, count value n=n+1; wherein M is a set width proportion coefficient; (7) Compare the count value n with the H value of the three-dimensional data of the wood. If the value of n is greater than N*H, it is determined that there are multiple abnormal data of wood pieces; where N is the set abnormal contour ratio coefficient. (8) The following are the specific steps for the secondary judgment of the three-dimensional data of wood that are detected as normal: (8.1) removing D profile data at the beginning and the end of the wood three-dimensional data to obtain new three-dimensional profile data W i , i = D, D + 1, …, H - D} (8.2) Calculate the total number of valid values P of each profile data i , {i = D, D + 1,..., H - D} of the quadratic gradient value Y i = {[(P i+1 -P i ) / 2]-[(P i+3 -P i+2 ) / 2] / 2}, {i = D, D + 1,..., H - D}; (8.3) Set initial count value m = 0 and gradient threshold K, when the second gradient value Y i , {i = D, D + 1,..., H - D} is greater than the set threshold K, count value m = m + 1; (8.4) Compare the cumulative value m with the set threshold T. If m>T, it is determined that there are multiple abnormal data of wood pieces. Otherwise, the three-dimensional data of wood is normal data.

Images