Small UAV infrared targets detection based on fourth-order central moment and local contrast feature fusion

To address the high false alarm rate of small unmanned aerial vehicle (UAV) infrared targets detection under complex backgrounds, this study presents the fourth-order central moment and local contrast features fusion to solve the above problems. Initially, infrared images are filtered by improved high-boost filtering (IHBF) to reduce most background noise and highlight potential target pixels, and the gray-level difference of the filtered image is improved through the square operation. Subsequently, a sliding window is applied to traverse the entire image, separately calculating the fourth-order central moment and local contrast to generate their respective feature maps. Then, an operator based on the fusion of the fourth-order central moment and local contrast features was defined to represent the gray levels of different regions of the preprocessed image, enhance the target and suppress the background. Finally, the extraction of the target is achieved through the adaptive threshold segmentation algorithm. The experimental results in five real-life scenarios show that, compared with other similar techniques, our method occupies the best detection index.