To achieve high- precision passive ranging of binocular CCD over a long distance, two algorithms of cross-correlation matching algorithm and cubic surface fitting algorithm are put forward, to match two pictures by binocular CCD in a sub- pixel level. The ranging principle of the binocular ranging system is introduced, and the ranging images are matched and a grayscale cross- correlation algorithm is proposed based on template.According to the results of cross-correlation, the sub-pixel level matching algorithm that uses the cubic surface fitting method to fit the sub- pixel coordinates of correlation peaks is proposed. The errors of the experimental results are analyzed and the formula of calculating system errors are corrected. Experimental results show that the actual measurement accuracy of a target within 3 km is better than 0.5% , which meets the accuracy requirements of the high-precision binocular ranging.