Center For Space Science And Applied Research Chinese Academy Of Sciences

Planetary rovers uniquely benefit planetary exploration; they enable regional exploration with the precision of in-situ measurements, a combination impossible from an orbiting spacecraft or fixed lander. The visual system belong to the onboard perception system, which is the main part of planetary rover. Visual information processing technology is very important for Rover vision system. But there is a still gap about it between our country and advanced one. Consequently, it is meaningful for theory and application to do deeply research on information processing technology. This thesis first analyzes the key technologies in visual navigation system, then projective rectification algorithm, stereo matching algortihm, obstacle avoidance algorithm, real-time image compression implementation in vision system are deeply researched.In chapter 2, a projective rectification algorithm for the vision system of planetary rovers is proposed. In this algorithm, the error function constructed by Sampson Error is minimized by Levenberg-Marquardt algorithm using corresponding points. The initial values are parameters calibrated on the earth. Then the parameters in the error function are simplified and restrained to confirm that the rectification transformation is an affine transformation from the optical point. Experiments show that the results are remarkably close to the ones produced by Euclidean rectification.Chapter 3 presents a low complexity and high performance stereo matching algorithm called Hardware Friendly Adaptive Support Weight (HFASW) and its FPGA implementation. This algorithm trades off the hardware implementation complexity and matching performance by using Census transform, quantizing the exponential function, two pass aggregation. The experiment results indicate that this algorithm and its FPGA implementation have high performance at processing speed and matching efficiency. The FPGA implementation can process 26 frames stereo images per second at 100MHz, and the average error rate is very low. The results of stereo matching are expressed by 3D reconstruction. In chaper 4, we build the 3D local terrain map using post-processed disparity map. And the experimental results indicate the terrain map can be used to obstacle avoidance. A stereo vision based rough terrain traversability analysis method called combined fitting plane/triple-line model for an autonomous planetary rover is then presented. Compared with the fitting plane model, the combined fitting plane/triple-line model can represent the terrain patch accurately when the point set in the terrain patch actually describes a non-planar surface. The proposed model has been applied to real 3D data with promising results though the input 3D data are sparse and of varying accuracy.For the application of image compression for the vision system, chapter 5 presents a high speed and high performance image compressor. The hardware solution proposed exploits a modified CCSDS algorithm and the processing speed is improved 40 times by using parallel architecture and pipeline technique. The experimental results indicate that the compression core has high performance at coding efficiency, data rate and error containment. A data rate of about 20Mpixels/s can be sustained at 50MHz.