Human detection and segmentation in the crowd environment by coimbining APD with HLBD approaches

The video-surveillance systems are popularly used in crowd monitoring and people detection in that crowd is estimated for security of humans in public places and also managing the resources. Humans can be detected and segmented through videos and this is very much important for security purposes.Today, there are many challenging problems in video surveillance due to variations in poses, clothing, lighting, illumination, occlusion and complexity of the background. The challenges in a complex background for video surveillance can be prevented in the human detection by using a model based approach.The template based shape matching algorithm is used to matching the different shapes of human.This approach is used to matching the boundary of the human shapes with detected image. Segmentation of the humans in a crowd scene takes both advantages of local region based detector and global shape boundary detector methods.Some of part based pose detector failed to detect different poses in the crowd environment.The proposed approach, different poses of human can be extracted by using adaptive pose detector.The different poses can be described by high level boundary descriptors.Normalized histograms are used as feature descriptors to find the optimal poses.By combining both APD and HLBD approaches, the misdetections of human are prevented in crowd environment. To classify human or non-human pattern in a crowd image sequence by using multiclass-SVM classifier. The proposed method is validated and evaluated using two public pedestrian dataset such as INRIA and crowded sequences, CAVIAR Benchmark and Munich Airport Datasets.