Positional Research Of A Quadruped Robot Based On Embedded Vision

The article mainly studies the visual positional issues during the task of a quadruped robot followed people. In the task, getting the distance from the quadruped robot to the target and angle from the target is the key to achieve robot followed target. The distance and angle deviation can be obtained by visual positioning. But achieving visual positioning of the quadruped robot needs do much of image processing in airborne system. For pipelined operating mode and strong logical computing power, FPGA can accelerate the visual system's speed of image process and is chose to be the chip of image processing. A quadruped robot vision positioning system is built based XC3SD3400A-4 chip and a visual positional algorithm is derived. The effectiveness of visual positional system is verified through experiments. The main work is completed as follows:(1) A brief mission requirement for the quadruped robot is described. Combined with specific relations of coordinate, the projection relationship's mathematical expressions of the camera are given and visual positional program is designed to target based on the target of visual positioning(2) For harris algorithm corner extraction unstable in the conditions of light changes, a proposed fixed-step search and divided by two strategies are proposed, improving the stability of the corner point extraction. Then the visual positional algorithm is derived based on perspective projection model of camera. Combining with inertial measurement unit(IMU) angle data, an effective correction scheme for positional algorithm is proposed. And the feasibility of positional algorithm is verified by experiments.(3) According to the positioning needs of the quadruped robot, vision positional system is built including camera data transmission module, image processing module, position estimation module. And the algorithms of image processing and positioning presented in Chapter III is realized in the system. Compared with the modules' time budget working in PC and in positional platform, the visual positioning platform is demonstrated that enables accelerating the speed of image processing.(4) In the conditions of depth and angles change, positional experiments are done by the visual positional platform. Experimental results show that the structures embedded vision positioning system can be achieved centimeter-level positional precision and the relative position and the deviation angle of the robot can be obtained by formula.