Research On Non-Color Feature Extraction Of Human Body For The Criminal Investigation

In recent years, with more and more perfect construction of video surveillance. Video information has been increasingly applied in investigation of crimes criminals. However, during the inspecting process by video surveillance, information of color and light is ineffective in most cases. That's because, on one hand, most criminal offenses take place at night and videos shoot at night have quite serious color distortion. On the other hand, most criminals have quite strong Counter investigation ability. Before and after the implementation of crime, they would change clothes, shoes&socks and hairstyle, which results in that extracted non-color features cannot assist in criminal investigation;. Due to its unchangeable property and non-compulsivity, extracting and detecting object's non-color information by video has become an urgent problem to be resolved by the security department. This paper has put forward a method of automatically extracting human's height based on single camera. Moreover, for increasing accuracy and extracting other features, it converts height extraction of a single picture to video extraction. Finally, by taking height information as the scale, it extracts other characteristics of measurement objective, main work of this paper is as follows:(1) It put forward the algorithm of automatic height extraction based on monocular camera. This algorithm is based on two reasonable assumptions. Meanwhile, for realizing automatic measurement, this paper also applies DPM algorithm and edge detection algorithm to calculate the metric points needed to be extracted. The simulation experiments show that even under the circumstance of unsatisfactory camera resolution and bite rate, the method can still achieve good results.So this algorithm can be better applied in actual video surveillance.(2) It converts the picture's automatic measurement to that of video. This paper put forward a method to eliminating mistakes and a method to reduce errors by periodicity. According to the experiment, it proves that both methods can better reduce mistakes and errors.(3) It extracts other characteristics by height information. This paper has put forward a method to extract gait periodicity by taking advantage of height changes. Meanwhile, it takes height of measurement objective as the scale, which can calculate head length of human body.It also means that, non-color information which can be extracted is not completed limited to information extracted in this paper,which lays a foundation for successive research.