Research On Labeling Method Of Transient Electromagnetic Radiation Source Based On Visual Information

The pulse and broadband signals generated by the transient electromagnetic radiation source are easy to produce potential electromagnetic interference threats to the frequency-using electronic equipment,which in turn affects its working performance and even causes potential safety hazards.In addition,the high-frequency components in the transient signal may also interfere with the device,resulting in communication errors or data loss.Therefore,it is of great significance to carry out the investigation and traceability of transient electromagnetic radiation sources,to explore the formation mechanism of electromagnetic interference,and to realize electromagnetic threat warning and electromagnetic safety protection.In this thesis,combined with computer vision technology,the visual annotation method of transient electromagnetic radiation source is studied by fusing visual information.The main research contents and innovations are as follows :(1)A monocular camera projection model based on camera pose estimation is established.By solving the camera internal parameter matrix and pose parameters,the monocular camera projection model is constructed,and the three-dimensional space coordinates of the target source are converted into scene image coordinates to realize the visual annotation of the transient electromagnetic radiation source.Through simulation test and overlap rate calculation,the main factors affecting the accuracy of labeling results are compared and analyzed,and experimental tests are carried out.The results show that the circular labeling box with a radius of 10 cm is used to visualize the high-voltage discharge power supply at two different positions.The overlap rates obtained by testing and calculation are 0.67 and 0.72,respectively.The labeling results have good directivity to the area where the high-voltage discharge is located.(2)A mapping labeling method for transient electromagnetic radiation sources under multi-view conditions is proposed.The SIFT feature points of the overlapping parts of different views of the binocular camera are used to solve the essential matrix between the cameras and their relative pose relationships.The genetic algorithm is used to optimize the scale information in the relative pose relationship of the binocular camera,and the sharing and association relationship of spatial position information under multi-view conditions is accurately constructed.The test results verify the feasibility of the method.(3)A visual labeling test system for transient electromagnetic radiation source is constructed.Based on the quaternary time difference positioning array and industrial camera,combined with the human-computer interaction interface developed by Python programming,a visual labeling test system was constructed to realize the fusion processing of scene visual information and electromagnetic radiation information.Under the condition of multi-view,the visual labeling test of transient electromagnetic radiation source in static scene and dynamic scene is carried out.The results show that the labeling overlap rate of the test system for the target source is greater than 0.5,and the labeling results intuitively show the area and location of the target source,which verifies the feasibility and effectiveness of the test system.