| Target detection in complex scenes is of great significance for the security and early warning of important areas such as power transmission channels and outdoor military bases.However,due to the complexity of the scene and high background noise interference,existing target detection technologies generally have problems such as single sensor types,inability to effectively identify targets,low detection efficiency,and lack of target positioning methods.Although the method of multi-sensor fusion compensates for the shortcomings of a single sensor,there are still problems such as difficulty in matching sensor data from different dimensions and poor adaptability to environmental changes.In this regard,this article combines sensor fusion technology with coordinate matching spatial models to conduct research on target detection methods for radar and video fusion,which is of great significance for improving the stability of target detection in complex scenes.Firstly,based on the perception mechanism of radar and camera,this paper designs a target detection framework based on radar and video fusion,and proposes a complex scene data fusion method based on improved frame difference to solve the problem of large noise interference and difficult detection in complex scenes.By adding an empirical factor to the frame difference method as a weighted average grayscale coefficient,the contrast between complex scenes and targets is highlighted,and the image where the target is located is clustered using contour extraction and scene coverage methods to achieve accurate detection and localization of complex scene targets.Further addressing the issue of missed and false detections caused by significant differences in spatial fusion,a fusion network based on the Fast R-CNN infrastructure is proposed.By correlating the detection results of consecutive moments as the next moment,a candidate region with prior information is provided for the current moment’s detection,avoiding association failure caused by accuracy differences.At the same time,the distance and velocity information measured by radar is used to provide the physical attributes of the target to obtain more accurate target results.The experimental results show that while accurately detecting complex scene targets,this method can effectively reduce the rate of missed detections and false detections. |
