Car Electronic Image Stabilization Technology

The study of the algorithm of image stabilization has an important effect in the realization of EIS. This paper introduced the basic concept and principle of electronic image stabilization technique and several algorithms in common use. According to the characteristics for vehicle-borne image process, analyzes the effect about the unstabilization of vehicle-borne image sequences to the result which observed by the vehicle-borne system image,and inject the concept of vehicle-borne electronic image stabilization technique. The implement of gray projection matching algorithms applied to stabilization of image sequences taken by vehicle-borne imaging sensor is introduced. This system adopt Power PC7410 as DSP image processor which including four CPU to realized "real-time" processing to vehicle-borne image sequences in the VxWorks OS.In the vehicle-borne imaging system, use the maximum gray level of the characteristic image to estimate the motion of image sequences between frames. In order to guarantee the validity of this algorithm, firstly the characteristic of an image is extracted. Then the frame is divided into several blocks. In each block feature points are chosen to estimate the motion vectors according to certain scanning order. At the same time, the camera is fixed on a plate form scanning over the relevant space area for catching moving target. Combination the characteristics of the vehicle-borne image tracking system are working together with the servo control system, amalgamation the scan movement which is the servo-tracking system detected the goal movement as a correction factor to deal with electronic image stabilization feedback from the scan movement in order to enhance the accuracy of electronic image stabilization.A lot of image stabilization experiments have been carried out with the video image sequence taken by vehicle-borne imaging system and the stabilization effect is evaluated. The experimented results show that this algorithm can realize image registration on vehicle-borne accurately and quickly with the registration precision up to"sub-pixel"level which means the processing capacity reached 50 fields per second, and the RMSE is smaller than 0.4. This capacity completely satisfy the demand of real-time and precision for image tracking system on vehicle-borne.