Tensor recovery infrared weak small target detection method combined with ATV constraint
A technology of weak and small targets and detection methods, applied in image data processing, instruments, calculations, etc., can solve the problem of high false alarm rate
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0111] Such as Figure 1-15 As shown, a tensor recovery infrared weak target detection method combined with ATV constraints includes the following steps:
[0112] Step 1: Construct the third-order tensor of the original image;
[0113] Step 2: Extract the prior information of the original image and construct the prior information weight tensor;
[0114] Step 3: Utilize tensor logDet function and tensor l 1 Norm, combined with ATV constraints, constructs an objective function, inputs the third-order tensor and prior information weight tensor into the objective function, and uses ADMM to solve the objective function to obtain the background tensor and target tensor;
[0115] Step 4: Reconstruct the background image and target image according to the background tensor and target tensor;
[0116] Step 5: Carry out adaptive threshold segmentation on the target image to determine the position of the target, and output the target detection result.
[0117] In order to improve the ...
Embodiment 2
[0120] Based on embodiment 1, the steps of this application are refined, and the technical means adopted for solving technical problems are described in detail: using tensor logDet function and tensor l 1 Norm, combined with ATV constraints, constructs an objective function, inputs the third-order tensor and prior information weight tensor into the objective function, and uses ADMM to solve the objective function to obtain the background tensor and target tensor.
[0121] Step 1 includes the following steps:
[0122] Step 1.1: Acquire the infrared image D∈R to be processed m×n , with a size of 245×326;
[0123] Step 1.2: Use a sliding window w with a size of 40×40 to traverse the original image D with a step size of 40, and use the matrix with a size of 40×40 in each sliding window w as a frontal slice;
[0124] Step 1.3: Repeat step 1.2 according to the number of window slides (63 in this example) until the traversal is complete, and form a new third-order tensor with all f...
Embodiment 3
[0176] Based on embodiment 1, this embodiment refines step 2, extracts the prior information of the original image, constructs the prior information weight tensor, uses the prior information related to the background and the target, ensures that the target is not distorted, and speeds up the algorithm The convergence speed also improves the robustness of the algorithm.
[0177] Step 2 includes the following steps:
[0178] Step 2.1: Define the structure tensor of the original image D J ρ It is defined as follows:
[0179]
[0180] Among them, K ρ Indicates the Gaussian kernel function with variance 2, * indicates the convolution operation, D σ Indicates that the Gaussian smoothing filter with a variance of 9 is performed on the original image, represents the Kronecker product, Indicates the gradient, means D σ the gradient along the x direction, means D σ Gradient along the y direction, J 11 replace J 12 Substitute K ρ *I x I y , J 21 Substitute K ρ *...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More