Information Fusion Technology On Multisensor And Its Application And Research In Tracking

The main purpose of this paper is to implement a traffic incidents detection system of video. The principium of vehicle detecting and tracking and the application of filter technology and multi-sensor image data confusion technology in tracking are researched. A experimental prototype system is designed and implemented, which can be used to work as a platform for the further research. The main contents are as follow:(1) A traffic detection algorithm combined of background subtraction algorithm and frame-by-frame subtraction algorithm is designed. The split of the vehicle profile got by using background subtraction is large. The vehicle profile got by using frame-by-frame subtraction is divided into several parts. The vehicle profile is obviously improved got by using the combination of these two methods. The experimental result shows that this method has a good effect on getting the vehicle profile.(2) A kalman multi-target tracking algorithm in single-sensor is designed. The kalman filter technology is used to predict and narrow the matching area. The characteristcs for example the vehicle location,profile area and the distance from the heart got during the detecting process are used to implement the vehicle tracing. It can meet the need of real time and veracity.(3)A experimental platform of binocular collection and confusion of image information is set up. Preprocessing including the adjustment of the frame rate and time synchronization must be done to the video before the confusion has been done because of the large difference of the two sensors. During the confusion process, the information got by the two sensors must be transferred to the same scene space. The corner matching algorithm and affine transformation are used in this paper. Meanwhile the necessary difference process is done to the affine transformation. A complete scene transformation model is constructed and it works well. (4)The vehicle information confusion tracking in double-sensors is implemented. The double-sensors video image information confusion technology is introduced to solve the deviation of profile extracting,large difference of the distance from the heart caused by single sensor tracking. The two sensors are used to track with kalman at the same time, and the vehicle information got by the two sensors is transferred into the same scene, then the characteristics information are confused. The optimal vehicle information would be got by doing the kalman filter again.