COUNTING OF PEOPLE IN A CROWDED IMAGE

Author’s Name : Amit Joshi | Chetan Zade | Divyank Shukla | Meet Bhimjiyani

Volume 02 Issue 04  Year 2015  ISSN No:  2349-3828  Page no: 1-3

Abstract – Crowd counting and management are a area of quite interest in field of computer vision. The operation of crowd counting can be useful in many filed like security, statistical researches. The data collected from crowd counting will also be helpful for further management of the crowd,.This paper uses mainly Head and shoulder parts of the human body to detect the presence of the human in a given region of interest. Instead of using a single detector to detect the images, use of multiple part detectors is proposed, which are learned by a boosting method. The results from all the part detectors are combined together to give out a single results i.e an estimated crowd count. Use of part detectors also helps tackle the problem of crowd occlusion to some level.The main advantage of part detectors is that even though complete human face and structure are not visible, they can still detect most of the partially visible parts.There are many papers for the purpose of humans detection and counting, but most of them are for videos. They also don’t deal with added problems like occlusion. Using the part based approach, we try to address this problem.

Keywords –Computer Vision,Crowd counting,Image Processing