This paper proposes a new feature-based method for tracking moving objects under the influence of disturbances that occur often in the real environment. Feature-based matching frequently suffers accuracy and performance due to the high number of image features that need to be matched. The suggested technique is based on corresponding color segments. Unlike other approaches, the generation of color segments is accompanied by segmentation of moving objects. Using segmentation of moving objects, it is possible to substantially reduce the probable set of image features that may potentially match. Thus, a high robustness is reached in strongly disturbed real image sequences.
展开▼
