A small target detection method for long-distance UAV based on infrared image

Abstract

A method and system for detecting a small target of a long-distance UAV based on an infrared image provided by the present invention firstly constructs the cumulative directional derivative convolution kernel of multiple scales in each direction of the infrared image; then the output of the average absolute gray scale difference algorithm Optimize, and then use the cumulative directional derivative convolution kernel to calculate the weighted coefficient of the optimized average absolute gray difference algorithm for each scale, obtain the weighted coefficient of each direction on each scale, and use the weighted coefficient of each direction on each scale to The smallest response is taken as the response map of this scale. Secondly, the response map is used as the weighting coefficient of the optimized average absolute gray difference algorithm on each scale to obtain the response maps on different scales. Finally, the maximum response map on different scales is used as The final response map is to extract the outline of the final response map to obtain the suspected target on the infrared image. This method effectively suppresses the background clutter, enhances the target area, and realizes accurate detection of the target.

Description

technical field [0001] The invention relates to the field of unmanned aerial vehicle detection, in particular to a long-distance unmanned aerial vehicle small target detection method based on infrared images. Background technique [0002] Infrared Target Search and Tracking System (IRST) is still difficult to detect long-distance small targets in infrared images due to the characteristics of concealed operation and anti-electronic interference due to low signal-to-noise ratio and complex background environment interference. Detection research can be divided into two categories. The first category is based on background removal methods. In the first category, desired operators (such as maximum median value, morphological opening operation, principal component tracking) are used to learn approximate background. Then, the approximate background is subtracted from the original image. Finally, an appropriate threshold segmentation strategy is used to extract objects. Although t...

AI Technical Summary

Problems solved by technology

Although this method is straightforward and has low computational complexity, it has a high false alarm rate for complex backgrounds.

And the camera can only point to a fixed area, cannot scan the learning background, and the detection range is limited

[0003] The second type of method is to directly enhance the target area and suppress the background area in the process of infrared small target detection. Considering the Gaussian model of the spatial distribution of small targets, the normalized Laplacian of Gaussian (LoG) is used in the scale space to solve the continuous problem. The change of the target size in the infrared image, although this method can enhance the low signal-to-noise ratio (psnr) target, it is sensitive to background noise due to the use of the second derivative

[0004] However, since the mean absolute gray scale difference algorithm has no difference between positive and negative contrast, the target area of ​​local contrast and the cavity area of ​​local contrast are both negative areas will be strengthened, such as figure 2 As shown, the area corresponding to the hole-shaped area is wrongly enhanced, which in turn increases the false alarm rate. At the same time, when the background of the infrared image contains high-intensity sharp edges, the detection performance of the AAGD algorithm drops sharply, as shown in image 3 shown

Images