Key Technology Research On Image Guidance Tracking In Dynamic Smoke Environment

As an important passive guidance method in modern warfare,the optical imaging guidance system has the characteristics of high strike accuracy and strong destructive effect,which can directly affect the shape and process of the battlefield.However,in the actual battlefield environment,the presence of large amounts of smoke and dust can produce absorption and scattering effects on the infrared radiation of the target,which in turn interferes with the detection path of the guidance system and seriously affects the detection capability of the infrared imaging guidance system.To address these problems,this paper investigates target detection and target tracking technology in dynamic smoke and dust environments from the perspective of smoke and dust interference characteristics.The main research contents are as follows:(1)Based on the working principle of infrared imaging guidance system to analyze the interference effect of smoke and dust,the causes of battlefield smoke and dust generation,particle characteristics and spatial distribution characteristics are studied;A soot extinction model was established,a preliminary analysis of the extinction performance of different soot components was completed,and the variation curves of transmittance and guide head signal-to-noise ratio with the physical properties of soot were simulated;Finally,an experimental platform was built to verify the interference characteristics of soot on infrared imaging.(2)To address the interference of the soot environment on the infrared image enhancement processing,an infrared soot image enhancement method based on guided filter image layering is proposed,where secondary layering is performed by anisotropic diffusion,different enhancement strategies are used for each information layer image,and finally merged using a fusion strategy based on the average luminance.Compared with the original algorithm,the average gradient and information entropy of this algorithm in the two sets of scenes are increased by 3.91 and 1.16 on average.(3)For the effect of dynamic environment on motion target detection algorithm,the traditional Vibe algorithm is improved by adding dynamic adaptive thresholds and adjusting the model update strategy to enhance the robustness of the detection algorithm under dynamic background.Compared with the traditional Vibe algorithm,the method in this section achieves an average improvement of 29% in recall,23% in precision,and 27% in F-value for the three dynamic background scenes.(4)The infrared video dataset was built and the ground truth of each video sequence was manually annotated for analyzing the tracking algorithm of infrared images.An improved KCF target tracking algorithm combining occlusion determination and Kalman trajectory prediction is proposed for the occlusion problem of KCF algorithm in dynamic smoke environment.With the comprehensive data set,the algorithm in this paper improves 15.2% in terms of accuracy and 10.9% in terms of success rate compared to the KCF algorithm.The improved algorithm also maintains a high overlap rate in the soot masking test sequence,which can better adapt to partial masking and short-time complete masking in the soot environment.