Research Of Infrared Target System Based On DSP

Target tracking has always been a research hotspot in the field of computer vision.As the target tracking of visible light is often lost during the night or when it is affected by smoke,the infrared image is less affected at night,so the infrared target tracking is obtained.The wide range of applications are used militaryally for weapons guidance,drone reconnaissance,etc.,and are used in everyday life for transportation,video surveillance,and so on.However,since the infrared image texture features are not obvious,it is difficult to track when there are problems such as noise and similarity interference.In order to solve these problems,this paper presents a multi-feature fusion based on DSP particle filter infrared target tracking algorithm.The main research work and innovations of the paper are as follows:1.There are few features that can be extracted from infrared targets.From the point of view of different features of pedestrian images with different characterization capabilities,this paper extracts the color features,edge features and texture features of infrared images,and fuses the three features.The effect of multiple feature fusions is that the feature effects thus obtained are more representative of a target.2.The effect of Meanshift filtering is greatly affected by similarity interference.Kalman filtering is applicable to linear,discrete and finite systems and the target will be lost after a long occlusion time.The particle filter method is suitable for many In the system,the feature is weighted by the likelihood function into the particle filter flow,which simplifies the operation flow of the particle filter and improves the tracking performance.3.Due to the adoption of multi-feature fusion method,there are many features extracted and the real-time tracking performance of the algorithm is low.The assembly function has the characteristics of high speed and high efficiency.Not high functions are processed to improve the efficiency of the algorithm.