Research Of UAV Detection Algorithm Based On Video

In recent years,the development of UAV technology has been changing with each passing day.UAVs are playing an increasingly important role in photography,surveying and mapping,agricultural remote sensing,logistics,flaw detection,and cleaning.However,drones have also caused serious problems in social life in some aspects.With the increase in the number of drones,the supervision of drones is imperative.Based on the above background,this paper presents a video-based UAV detection method,and designs the FPGA implementation of this method.At first this paper introduced the basic knowledge of moving target detection,and focused on GMM and Vi Be algorithm which effect better than other existing methods.Afterwards,experiments under different scenarios were designed to test the performance of the GMM and Vi Be algorithm.The GMM can fit the probability distribution of gray values of pixels in a complex background,but it cannot detect weak targets such as drones.Vi Be’s detection of weak targets is more sensitive than that of GMM,but the detection results in complex backgrounds have a lot of noise.In addition,the Vi Be algorithm requires only one frame of image to initialize the background model.The update strategy is simple and the amount of computation is small,which is easy to implement using an embedded system.So this paper combines the advantages of the GMM algorithm,improves the Vi Be algorithm,and proposes a video-based drone detection method.This paper improves the Vi Be algorithm mainly in the preprocessing and foreground detection stages.In the preprocessing stage,a horizon detection method is proposed,which divides the sky and ground regions according to the distribution of the gray values of the images in each column.This segmentation method can avoid the interference of objects on the ground to the detection of the drone,and can also reduce the computational complexity of the algorithm.In the foreground detection stage,combined with the core idea of GMM,a Gaussian distribution is introduced during threshold detection.Calculate the distance between the pixel gray value and the sample mean,and use the 2.5 times standard deviation of the sample as the threshold to segment the foreground and background.This makes the algorithm work better in complex backgrounds.After experiments,the effectiveness of this method in the detection of drones in a complex background is significantly higher than that of GMM and Vi Be.In the end,according to the characteristics of FPGA,this paper makes a targeted adjustment to the algorithm and designs an FPGA implementation method for each step of the algorithm.And through simulation,it is verified that the FPGA module designed in this paper can implement this method in real time.