Taras P. Riopka and Peter Bock
Current approaches in motion analysis often rely on differencing operations that inherently amplify noise and are hindered by the spatial correspondance problem. An alternative approach is proposed using ALISA (Adaptive Learning Image and Signal Analysis) to detect differences in types of motion by analyzing single frames. Using an appropriate set of features, the ALISA engine accumulates a multi- dimensional histogram that estimates the probability density function of a feature space, and uses the result as a basis for classification. As a function of image sampling rate and the scale of image structures, the ALISA engine was able to discriminate between a slow moving and fast moving object with a confidence exceeding 99 percent.