Abstract
A new normalization is employed to increase the reliability of correlation method for a distance measurement of a target, for instance grass-land, which has random and poor texture with low contrast of brightness. It is clarified that gain and bias-adjustment at A/D conversion of a stereo-pair pictures arise a problen to mislead the correlogram, and is shown theoretically that an adaptive normalization of the pictures is effective to solve the problem. A normalized cross-correlation (NCC) to obtain homologous points in the stereo-pair is derived by the new normalization. A practical methed employing a fast algorithm to calculate the NCC in real time is proposed, and it shows that the normalization time is independent of the number of data. Finally, the present method is applied to measure the horizontal distances using triangulation at 25 points on the landing-strip of the Miyanomori jump hill, and the profile of the landing-strip is reconstructed by result of the measurements. The specification of the experimental set-up is specially cared in order to realize a handy type range finder with real time ability by the present method. Data processing is carried on a mini-computer and results of the experiment show that the NCC is useful to find the true homologous points in noise-like-texture, and that the fast algorithm is adequate for real time data processing because the execution time of the NCC by the algorithm is one-tenth or less than that by ordinary methods.
