Dual-core DSP based stereo vision positioning of point targets

Four positioning algorithms are discussed as follows: homogeneous method, nonhomogeneous method by solving the normal equation directly, nonhomogeneous method by generalized inverse, analytic geometry method. A dual-core DSP parallel and interactive cross work is proposed to realize 2D position correction and 3D positioning. Experiments are carried out to investigate the real time and accuracy of the algorithms. Experimental results show that when a frame of image with four point targets are processed by the nonhomogeneous method by solving the normal equation directly, the latency between the original data and the positioning results are 9.21ms, and the accuracy is 0.15mm RMS, which shows satisfactory real time and accuracy performance.