A UAV aerial vehicle detection method based on target spatial distribution characteristics

Abstract

The invention discloses a vehicle detection method for aerial photography of unmanned aerial vehicles based on target spatial distribution characteristics. The method firstly constructs an aerial vehicle density estimation network based on the idea of ​​conditional generation against CGAN, generates a vehicle density map, and thereby obtains the spatial distribution characteristics of vehicle targets; Secondly, according to the spatial distribution characteristics of the target, the high-resolution aerial vehicle image is adaptively segmented to obtain several local image blocks; finally, a single-stage vehicle detector is obtained by using the local image block and the original UAV aerial vehicle image to train respectively, After the detection results of the global image and local image blocks are fused at the decision level based on the Soft-NMS algorithm, the final detection results are output. The UAV aerial vehicle detection method proposed by the present invention avoids the loss of target pixel feature points caused by the scaling of the original image, and further improves the detection accuracy of the vehicle.

Description

technical field [0001] The invention belongs to the field of intelligent transportation, and in particular relates to a vehicle detection method based on the spatial distribution characteristics of an unmanned aerial vehicle. Background technique [0002] In recent years, the research on intelligent transportation systems has attracted extensive attention from scholars from all walks of life. Accurate and real-time comprehensive perception of road traffic scene information is an important link in the construction of intelligent transportation systems. Among them, the detection and identification of the main traffic participants vehicles , is a necessary prerequisite for traffic situational awareness assessment. Compared with the way of installing fixed cameras on the roadside, the perception of vehicles in road traffic from the perspective of UAV aerial photography has the advantages of high flexibility, wide viewing angle and large range, which can make up for the environme...

AI Technical Summary

Problems solved by technology

However, the average segmentation method will lead to problems such as a certain image slice being all invalid background areas and truncation of vehicle targets, which will affect the accuracy of target recognition.

Images