基于Top-Hat变换与SVM的无人机红外目标检测OA

In view of the low contrast,high real-time property requirements,and limited embedded platform resources in infrared UAV target detection,this paper designs an embedded infrared target detection system based on Zynq XC7Z035 SoC.The system adopts an ARM-FPGA collaborative architecture,where the FPGA implements infrared video data acquisition,video direct memory access(VDMA)high-speed transmission,and video stream output,while the ARM runs a customized embedded Linux system integrated with a streamlined OpenCV library.A lightweight detection algorithm based on Top-Hat transform and support vector machine(SVM)classifier is developed.The algorithm enhances small target features and suppresses background interference by morphological Top-Hat transform.A linear kernel SVM is employed for target classification,and the detection stability is further improved by combining non-maximum suppression(NMS)and temporal filtering.The test results demonstrate that the system achieves a comprehensive detection success rate of 91.3%under various background conditions,while maintaining a real-time processing speed of 31 fps with guaranteed detection accuracy.This study provides an effective solution to balance recognition precision and computational efficiency in infrared target detection.