Target characteristic extraction algorithm based on block structure variables

摘要

In image processing system, the target recognition is very crucial. An optimized algorithm based on the definition of block is presented. In this algorithm, two structure variables are self-defined to calculate the areas and centroids of targets. The labeling conflicts are resolved by tracking and correcting, which means that if there is a conflict, trace the neighbored blocks and correct the labels of them. Finally, the characteristics of blocks, which have the same label, are accumulated. This method has outstanding advantages in saving memory, compared with Pixel labeling algorithm and Run-length code algorithm. The new algorithm is simple and the results of target characteristics are easy to be analyzed and handled in the following processes.