Research on object detection method in smoke and fog environment

Abstract

Laser fuse has the advantages of good monochromaticity, high collimation and strong anti-electromagnetic interference ability. However, when passing through smoke and fog media, laser fuse will interact with suspended particles, causing part of the energy to be scattered by suspended particles to an angle different from the target direction, part of the energy to be absorbed by particles and other complex effects. Thus, attenuation of beam energy and false alarm are generated, and detection performance of laser fuse is reduced. When the target is in the smoke and fog background, it is difficult to recognize that the target echo overlaps with the smoke echo. To solve this problem, the target echo signal is recognized based on the Gaussian decomposition combined with SVM (Support Vector Machine) algorithm. The Gaussian waveform decomposition method is used to separate the echo signal of smoke and target, and the amplitude, center position and half wave width of the two waveforms are calculated. SVM algorithm is used for training. The signals of smoke and targets at different distances in different concentrations of smoke environment are collected, and the experimental verification shows that the recognition accuracy can reach 91.2%, which is compared with the existing KNN, K-Means, SAE algorithms. The results show that SVM algorithm has excellent recognition effect in smoke environment with different concentrations.