IMPLEMENTATION OF MOVING OBJECT DETECTION AND CATEGORIZATION FROM HEVC COMPRESSED SURVEILLANCE VIDEO

Many important applications in medical field, monitoring activities in public areas and businesses or commercial buildings, identifying abnormal events in traffic are done using moving object identification from HEVC surveillance videos. In this paper, we are using a procedure that detects and classifies moving targets from videos compressed per HEVC standard. The first thing we are doing is pre-processing and it is done using moving vector consolidation for intra-coded prognosis unit and MV exception expulsion. The second step is to apply four-connectivity component labelling algorithm for clustering of obstructs with non-zero moving vectors into associated forefront locales. The third thing is to remove noise regions from connected foreground regions based on temporal consistency by object region tracing. Utilizing the coding unit size and prognosis unit estimate the further refinement of limit of moving item locale is ended. At last either people or vehicles i.e., moving targets are ordered from a person vehicle categorization prototype prepared utilizing pack of spatial-transient HEVC grammar words. Our experimental results categorize moving persons and vehicles accurately and robustly with remarkable performance.