Study On Key Technology Of Panoramic Image Stabilization Based On Vehicle-Mounted Platform

The infrared panoramic search system based on line detectors can achieve target search of 360 full airspace and recognition of threat targets,and has the advantages of passive detection,high reliability and strong anti-interference ability.Influenced by road bumps,video has a certain degree of jitter,which not only affects the visual effect,but also affects the accuracy of subsequent target recognition or tracking.Therefore,it is necessary to eliminate the vehicle vibration to image sequences by the image stabilization technique.Due to the advantages of small volume,light weight and easy operation,electronic image stabilization has become the main direction of development.The vehicle-mounted infrared panoramic search system based on line detectors is used as the research object in this paper,according to the imaging characteristics of this system,and the in-depth research on the key technology of electronic image stabilization under panoramic view is carried out.The specific research contents are as follows:(1)Introduction and analysis of the basic theory of electronic image stabilizationFirstly,the fuzzy reasons of the image sequence are analyzed,and the infrared imaging system,the imaging model of the camera,motion forms and two-dimensional image motion models are introduced.Secondly,the principle of image motion compensation and the basic structure of the electronic image stabilization system are introduced.Finally,the key theories of each link in the image stabilization technology are analyzed in detail,which provide theoretical support for the research and implementation of the following panoramic image stabilization algorithm.(2)Study of panoramic image stabilization algorithm based on inertial navigation informationIn the absence of stabilized platform,in order to achieve the goal of stable search in vehicle traveling,it is necessary to eliminate the effects of vehicle vibration on imaging results,ie,the rotational motion in the three directions of heading,pitching,and rolling.The panoramic image stabilization algorithm based on inertial information is proposed in this paper,which collects the three-dimensional attitude angle datas of the vehicle in real time through the inertial navigation device,and realizes the reverse compensation of the current frame image by gesture compensation model and image mapping relationship.The algorithm is easy to operate,fast,and not easy to be disturbed by the motion foreground,and can achieve good panoramic image compensation results.(3)Study on image stabilization algorithm based on adaptive compensationSince the panoramic image stabilization still needs to be transformed into the image stabilization problem of small field of view,it is particularly necessary to further improve the image stabilization accuracy of small field of view.Aiming at the drawbacks of motion estimation and motion compensation in traditional electronic image stabilization algorithms,the image stabilization algorithm is optimized and improved in this paper.Firstly,in order to further improve the matching accuracy and efficiency of the feature points,a feature point matching strategy based on clustering analysis is proposed,which combines the K-means clustering analysis principle to eliminate error matching points.The experimental results show that compared with the distance suppression constraint algorithm and the random sampling consistency algorithm,the matching algorithm in this paper can guarantee higher correct matching rate and shorter matching time,and has good robustness.Secondly,aiming at the problem of inter-frame hopping or cumulative error in traditional motion compensation methods,an image stabilization algorithm based on adaptive compensation is proposed,including preprocessing,feature point matching,adaptive update of reference frames,reference frames,cumulative iteration of global motion parameters,Kalman filtering,motion compensation and so on.The experimental results show that compared with the fixed frame compensation method or the adjacent frame compensation method,the video after the proposed algorithm has a higher peak-to-noise ratio between frames,and the inter-frame transition becomes smoother.(4)Study on the panoramic image stabilization algorithm based on region segmentation and fusionIn the panoramic image,due to the jitter difference of different columns and the inconsistency of imaging characteristics in the local area,the traditional image stabilization algorithms for small field cannot be directly applied.Therefore,a panoramic image stabilization algorithm based on regional segmentation and fusion is proposed,which transforms the panoramic view field image stabilization into the image stabilization problem in the small field of view.Firstly,pre-compensation is used to reduce the error of pitch base between columns and columns,and the panoramic image region is divided into the front region,the end region,the left region and the right region.Then,the local regions are respectively stabilized through different image stabilization models,and the fusion of the overlapped regions was combined with the asymptotically weighted average fusion method to ensure the smooth transition.The experimental results show that compared with the undivided panoramic image stabilization algorithm,the local detail processing of the algorithm is more delicate and the image stabilization accuracy is higher.