Detection and classification for group moving humans

Abstract

In the case of moving group of humans the recognition algorithms more often misclassify it as vehicles or large moving object. It is there fore the aim of this project to detect and classify moving object as either Group of humans or something else. The background subtraction technique has been employed in this work as it is able to provide complete feature of the moving object. However, it is extremely sensitive to dynamic changes like change of illumination. The detected foreground pixels usually contain noise, small movements like tree leaves. These isolated pixels are filtered by some of preprocessing operations; such as median filter and sequence of morphological operations dilation and erosion. Then the object will be extracted using border extraction technique. The classification makes use the shape of the object. The performance of the proposed technique has achieved 75% accuracy based on 18 test samples. This result shows that if it possible to distinctly classify a group of humans moving in the video sequence from other large moving objects such as vehicles.