A Robust Multi-Model Approach for Face Detection in Crowd

摘要: The estimation of the number people in surveillance areas is essential for monitoring crowded scenes. When density a zone increases to certain approximated level, people's safety can be endangered. Detection human prerequisite estimation, tracking, activity recognition and anomaly detection even non congested areas. This paper presents robust hybrid approach face crowd by combining skin color segmentation Histogram Oriented Gradients(HOG) with Support Vector Machine(SVM) architecture. Initially, image enhancement performed improve rate. An edge preserving pyramidal applied multiscale representation an image. Skin done combination YCbCr RGB model, HOG features are extracted from segmented region. We trained SVM classifier Muct FEI databases which consist 751 2800 images respectively. accuracy this evaluated testing it on BAO multiple database various manually collected captured Experimental results demonstrate that supplementary more potent increasing rate than using only. proposed achieves 98.02% higher comparison Viola Jones fast method.