High-dynamic multi-target unmanned aerial vehicle detection method based on feature fusion

Abstract

The invention provides a high-dynamic multi-target unmanned aerial vehicle detection method based on feature fusion, and the method comprises the following steps: carrying out the real-time detection, and obtaining an image; feature extraction and feature fusion; performing morphological operation; searching a target contour; searching a target; error compensation is carried out on multi-detection and leakage detection existing in detection; and outputting a target according to the judgment of detection and compensation. The method mainly aims at the problem that targets are easy to lose or mistakenly detect when a plurality of moving targets are tracked to move in a staggered manner, has certain robustness for illumination, extracts an original binary image of the target through color space conversion, and judges whether the filtered binary image is subjected to morphological expansion or not through brightness detection after median and Gaussian filtering; and then edge detection is carried out on the accurate target binary image to obtain a minimum surrounding rectangular frame, and after frame-by-frame detection, a motion track corresponding to the target corresponding to the color is drawn to complete multi-target video identification.

Description

technical field [0001] The invention relates to the technical field of dynamic monitoring of multi-target unmanned aerial vehicles, in particular to a high-dynamic multi-target unmanned aerial vehicle detection method based on feature fusion. Background technique [0002] The detection and identification of drones is an important challenge in security. In the military field, there is still a lack of defense means for drones. In practical applications, the ground contains a lot of clutter in the down-looking scene in ground and low-altitude radar target recognition. There are also some scholars who have used deep learning-based algorithms in their research, which require a large number of standard samples in the process of target detection and recognition. There are many kinds of drones of different sizes and shapes. Collecting all types and sizes of drones as training data sets for training is obviously a huge workload and difficult to achieve. The traditional UAV low-alti...

AI Technical Summary

Problems solved by technology

[0003] In view of this, the present invention provides a highly dynamic multi-target UAV detection method based on feature fusion, which is mainly aimed at the problem of easy loss of targets or false detection when tracking multiple moving targets interlaced, and has certain robustness to illumination , the target original binary image is extracted through color space conversion, after median and Gaussian filtering, it is judged by brightness detection whether it performs morphological expansion on the filtered binary image; then edge detection is performed on the more accurate target binary image, Get its minimum enclosing rectangular frame, and draw the motion trajectory corresponding to the color-corresponding target after frame-by-frame detection to complete multi-target video recognition

Images