A Distributed X-ray Image Detection Method Based on Image Complexity

Abstract

This invention relates to the field of X-ray nondestructive testing technology, specifically a distributed X-ray image inspection method based on image complexity driving. The method includes the following steps: Step 1: Image acquisition; Step 2: Image segmentation and processing complexity calculation: Dividing the X-ray image into multiple image regions and calculating the processing complexity of each image region; Step 3: Node construction: Constructing multiple distributed processing nodes; Step 4: Image region allocation: Assigning each image region to each distributed processing node, ensuring that the performance parameters of each distributed processing node match the processing complexity of each image region; Step 5: Node optimization; Step 6: Distributed processing; Step 7: Detection result generation. This method solves the problem of low detection efficiency in traditional X-ray nondestructive testing methods and is suitable for internal defect identification and performance evaluation in aerospace equipment, rail transportation equipment, energy equipment, electronic components, and porous media materials.

Description

Technical Field

[0001] This invention relates to the field of X-ray nondestructive testing technology, specifically a distributed X-ray image detection method driven by image complexity. Background Technology

[0002] X-ray nondestructive testing (NDT) is a nondestructive testing technique that uses X-rays to penetrate a workpiece and detect internal defects without damaging it. It is widely used in aerospace equipment, rail transportation equipment, energy equipment, electronic components, and the identification and performance evaluation of internal defects in porous materials. In practical applications, with the increasing size of workpieces, the growing complexity of material structures, and the ever-increasing demands for detection resolution, it is often necessary to acquire and process a massive number of highly information-dense X-ray images. However, traditional X-ray NDT methods suffer from slow image processing speeds due to uneven utilization of computational resources, resulting in low detection efficiency. Therefore, it is necessary to develop a distributed X-ray image detection method driven by image complexity to solve the problem of low detection efficiency in traditional X-ray NDT methods. Summary of the Invention

[0003] To address the problem of low detection efficiency in traditional X-ray nondestructive testing methods, this invention provides a distributed X-ray image detection method driven by image complexity.

[0004] This invention is achieved using the following technical solution:

[0005] A distributed X-ray image detection method based on image complexity-driven approach, comprising the following steps:

[0006] Step 1: Image Acquisition

[0007] X-ray images of the workpiece are acquired using an X-ray imaging system;

[0008] Step 2: Image segmentation and processing complexity calculation:

[0009] The X-ray image is divided into multiple image regions according to a preset scale, and the features of each image region are extracted. Then, the processing complexity of each image region is calculated based on the features of each image region, and a complexity score map is constructed based on the processing complexity of each image region.

[0010] Step 3: Node Construction:

[0011] Construct multiple distributed processing nodes and initialize the performance parameters of each distributed processing node;

[0012] Step 4: Image Region Allocation

[0013] Each image region is used as the ant colony in the ant colony algorithm, and the mapping relationship between the processing complexity of each image region in the complexity scoring graph and the performance parameters of each distributed processing node is used as the path of the ant colony.

[0014] Then, the ant colony algorithm is used to perform multiple rounds of path search to obtain the optimal path of the ant colony, which is the optimal mapping relationship between the processing complexity of each image region in the complexity score graph and the performance parameters of each distributed processing node.

[0015] Then, based on the optimal mapping relationship, each image region is assigned to each distributed processing node, so that the performance parameters of each distributed processing node match the processing complexity of each image region in the complexity scoring map.

[0016] Step 5: Node Optimization:

[0017] The processing parameters of each distributed processing node are used as the particle swarm of the particle swarm algorithm. The particle swarm algorithm updates the velocity and position of the particle swarm through individual optimality and global optimality, thereby obtaining the optimal processing parameters of each distributed processing node.

[0018] Step Six: Distributed Processing

[0019] Based on the optimal processing parameters, each distributed processing node processes the image region assigned to it.

[0020] Step 7: Generating test results:

[0021] The processing results of each distributed processing node are merged, and the detection results are generated based on the fusion results.

[0022] Furthermore, in step two, the features of the image region include: the gradient statistics of the image region, the texture entropy of the image region, and the gray-level variance of the image region;

[0023] The formula for calculating the processing complexity of an image region is as follows:

[0024] ;

[0025] In the formula: Indicates the first The processing complexity of each image region; Indicates the first Gradient statistics for each image region; Indicates the first Texture entropy of an image region; Indicates the first Gray-level variance of each image region; , , All are adjustable coefficients.

[0026] Furthermore, in steps three and four, the performance parameters of the distributed processing node include: the computing power of the distributed processing node, the current load of the distributed processing node, and the task response time of the distributed processing node.

[0027] Furthermore, in step four, allocating each image region to each distributed processing node specifically means: allocating image regions with high processing complexity to distributed processing nodes with high performance parameters, and allocating image regions with low processing complexity to distributed processing nodes with low performance parameters.

[0028] Furthermore, in step five, the processing parameters of the distributed processing node include: the denoising threshold of the distributed processing node, the segmentation stride of the distributed processing node, and the convolution kernel size of the distributed processing node.

[0029] Furthermore, in step seven, the generation process includes: a. performing threshold segmentation, edge extraction, and feature matching on the fusion result to generate candidate targets; b. performing connected component analysis and morphological constraints on the candidate targets to eliminate false targets; c. identifying and classifying the candidate targets according to preset detection rules and a pre-trained classification model; d. generating detection results in a structured form based on the identification and classification results.

[0030] Furthermore, in step seven, the detection results include: the location of the detection target, the type of the detection target, the size parameters of the detection target, the confidence level of the detection target, and whether the detection target is qualified.

[0031] Furthermore, in step seven, the detection result is output to a display terminal or storage module after it is generated.

[0032] Compared to traditional X-ray nondestructive testing methods, the distributed X-ray image detection method based on image complexity driving described in this invention achieves intelligent matching between image regions and computing resources (i.e., distributed processing nodes) by calculating the processing complexity of image regions and combining ant colony optimization and particle swarm optimization algorithms. This enables large-scale distributed image processing. Therefore, the computing resource utilization of this invention is more balanced, thereby significantly improving image processing speed and detection efficiency.

[0033] This invention effectively solves the problem of low detection efficiency in traditional X-ray nondestructive testing methods, and is applicable to the identification and performance evaluation of internal defects in aerospace equipment, rail transit equipment, energy equipment, electronic components, and porous media materials. Detailed Implementation

[0034] A distributed X-ray image detection method based on image complexity-driven approach, comprising the following steps:

[0035] Step 1: Image Acquisition

[0036] X-ray images of the workpiece are acquired using an X-ray imaging system;

[0037] Step 2: Image segmentation and processing complexity calculation:

[0038] The X-ray image is divided into multiple image regions according to a preset scale, and the features of each image region are extracted. Then, the processing complexity of each image region is calculated based on the features of each image region, and a complexity score map is constructed based on the processing complexity of each image region.

[0039] Step 3: Node Construction:

[0040] Construct multiple distributed processing nodes and initialize the performance parameters of each distributed processing node;

[0041] Step 4: Image Region Allocation

[0042] Each image region is used as the ant colony in the ant colony algorithm, and the mapping relationship between the processing complexity of each image region in the complexity scoring graph and the performance parameters of each distributed processing node is used as the path of the ant colony.

[0043] Then, the ant colony algorithm is used to perform multiple rounds of path search to obtain the optimal path of the ant colony, which is the optimal mapping relationship between the processing complexity of each image region in the complexity score graph and the performance parameters of each distributed processing node.

[0044] Then, based on the optimal mapping relationship, each image region is assigned to each distributed processing node, so that the performance parameters of each distributed processing node match the processing complexity of each image region in the complexity scoring map.

[0045] Step 5: Node Optimization:

[0046] The processing parameters of each distributed processing node are used as the particle swarm of the particle swarm algorithm. The particle swarm algorithm updates the velocity and position of the particle swarm through individual optimality and global optimality, thereby obtaining the optimal processing parameters of each distributed processing node.

[0047] Step Six: Distributed Processing

[0048] Based on the optimal processing parameters, each distributed processing node processes the image region assigned to it.

[0049] Step 7: Generating test results:

[0050] The processing results of each distributed processing node are merged, and the detection results are generated based on the fusion results.

[0051] In step two, the features of the image region include: the gradient statistics of the image region, the texture entropy of the image region, and the gray-level variance of the image region.

[0052] The formula for calculating the processing complexity of an image region is as follows:

[0053] ;

[0054] In the formula: Indicates the first The processing complexity of each image region; Indicates the first Gradient statistics for each image region; Indicates the first Texture entropy of an image region; Indicates the first Gray-level variance of each image region; , , All are adjustable coefficients.

[0055] In steps three and four, the performance parameters of the distributed processing node include: the computing power of the distributed processing node, the current load of the distributed processing node, and the task response time of the distributed processing node.

[0056] In step four, allocating each image region to each distributed processing node specifically means: allocating image regions with high processing complexity to distributed processing nodes with high performance parameters, and allocating image regions with low processing complexity to distributed processing nodes with low performance parameters.

[0057] In step five, the processing parameters of the distributed processing node include: the denoising threshold of the distributed processing node, the segmentation stride of the distributed processing node, and the convolution kernel size of the distributed processing node.

[0058] In step seven, the generation process includes: a. performing threshold segmentation, edge extraction, and feature matching on the fusion result to generate candidate targets; b. performing connected component analysis and morphological constraints on the candidate targets to eliminate false targets; c. identifying and classifying the candidate targets according to preset detection rules and a pre-trained classification model; d. generating detection results in a structured form based on the identification and classification results.

[0059] In step seven, the detection results include: the location of the detection target, the type of the detection target, the size parameters of the detection target, the confidence level of the detection target, and whether the detection target is qualified.

[0060] In step seven, the detection results are generated and then output to a display terminal or storage module.

[0061] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A distributed X-ray image detection method based on image complexity-driven approach, characterized in that: The method includes the following steps: Step 1: Image Acquisition X-ray images of the workpiece are acquired using an X-ray imaging system; Step 2: Image segmentation and processing complexity calculation: The X-ray image is divided into multiple image regions according to a preset scale, and the features of each image region are extracted. Then, the processing complexity of each image region is calculated based on the features of each image region, and a complexity score map is constructed based on the processing complexity of each image region. Step 3: Node Construction: Construct multiple distributed processing nodes and initialize the performance parameters of each distributed processing node; Step 4: Image Region Allocation Each image region is used as the ant colony in the ant colony algorithm, and the mapping relationship between the processing complexity of each image region in the complexity scoring graph and the performance parameters of each distributed processing node is used as the path of the ant colony. Then, the ant colony algorithm is used to perform multiple rounds of path search to obtain the optimal path of the ant colony, which is the optimal mapping relationship between the processing complexity of each image region in the complexity score graph and the performance parameters of each distributed processing node. Then, based on the optimal mapping relationship, each image region is assigned to each distributed processing node, so that the performance parameters of each distributed processing node match the processing complexity of each image region in the complexity scoring map. Step 5: Node Optimization: The processing parameters of each distributed processing node are used as the particle swarm of the particle swarm algorithm. The particle swarm algorithm updates the velocity and position of the particle swarm through individual optimality and global optimality, thereby obtaining the optimal processing parameters of each distributed processing node. Step Six: Distributed Processing Based on the optimal processing parameters, each distributed processing node processes the image region assigned to it. Step 7: Generating test results: The processing results of each distributed processing node are merged, and the detection results are generated based on the fusion results; In step two, the features of the image region include: the gradient statistics of the image region, the texture entropy of the image region, and the gray-level variance of the image region. The formula for calculating the processing complexity of an image region is as follows: ； In the formula: Indicates the first The processing complexity of each image region; Indicates the first Gradient statistics for each image region; Indicates the first Texture entropy of an image region; Indicates the first Gray-level variance of each image region; , , All are adjustable coefficients; In step four, allocating each image region to each distributed processing node specifically means: allocating image regions with high processing complexity to distributed processing nodes with high performance parameters, and allocating image regions with low processing complexity to distributed processing nodes with low performance parameters.

2. The distributed X-ray image detection method based on image complexity as described in claim 1, characterized in that: In steps three and four, the performance parameters of the distributed processing node include: the computing power of the distributed processing node, the current load of the distributed processing node, and the task response time of the distributed processing node.

3. The distributed X-ray image detection method based on image complexity as described in claim 1, characterized in that: In step five, the processing parameters of the distributed processing node include: the denoising threshold of the distributed processing node, the segmentation stride of the distributed processing node, and the convolution kernel size of the distributed processing node.

4. The distributed X-ray image detection method based on image complexity as described in claim 1, characterized in that: In step seven, the generation process includes: a. performing threshold segmentation, edge extraction, and feature matching on the fusion result to generate candidate targets; b. performing connected component analysis and morphological constraints on the candidate targets to eliminate false targets; c. identifying and classifying the candidate targets according to preset detection rules and a pre-trained classification model; d. generating detection results in a structured form based on the identification and classification results.

5. The distributed X-ray image detection method based on image complexity as described in claim 1, characterized in that: In step seven, the detection results include: the location of the detection target, the type of the detection target, the size parameters of the detection target, the confidence level of the detection target, and whether the detection target is qualified.

6. The distributed X-ray image detection method based on image complexity as described in claim 1, characterized in that: In step seven, the detection results are generated and then output to a display terminal or storage module.

Images