Convolutional neural network target detection and recognition method based on polarization scattering characteristics

Abstract

The invention discloses a convolutional neural network target detection and recognition method based on polarization scattering characteristics. The method comprises the following steps: carrying outirradiation and signal polarization on a target; an echo fed back from the target is received by means of dechirp, wherein the received echo is a target polarized echo; the target polarized echo is sampled, each frame of sampled radar echo data include echo data of a VV polarization channel, a VH polarization channel, an HH polarization channel, and an HV polarization channel; and the four-channelecho data are inputted by a convolutional neural network and thus the dangerous article on the target is detected and identified based on the target polarization characteristics. According to the invention, no imaging processing is required; the detection processing process is simple; the detection speed is fast; and the detection frame rate can reach 10Hz. With the convolutional neural network identification method, the fusion mode of the four-channel echo data with different polarization characteristics is found out, so that the article recognition rate is increased; and thus the certain detection and identification effect for dangerous articles under the body background is realized.

Description

technical field [0001] The invention relates to the technical field of human body security inspection, in particular to a non-imaging dangerous article detection and identification method based on target polarization scattering characteristics in the background of human body. Background technique [0002] The remote security inspection and early warning of human dangerous goods is a difficult problem of anti-terrorism in the world today. Due to its special frequency band, millimeter wave or terahertz has special properties such as penetrating clothing, so it has broad application prospects in the field of security inspection. [0003] At present, the stand-up human body security inspection mainly adopts the active imaging method to realize the detection of dangerous goods. The basic principle is to rely on the mechanical scanning structure to scan the target two-dimensionally to obtain the target image information. The mechanical scanning method greatly reduces its security...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a convolutional neural network target detection and recognition method based on polarization scattering characteristics, which can not only overcome the relatively complex technical defects in the signal processing process of the existing station-opening security inspection system, but also effectively solve the problem of slow detection speed. This method can play a certain role in the detection and identification of dangerous goods in the background of the human body, and the speed is fast, and the system structure requirements are simple

Images