Statistics On The Number Of People Indoor Based On Digital Image Processing

Counting the number of people in the Intercepted images from the video is a difficult and front subject in the field of image processing, It has high practical value in traffic monitoring, traffic statistics, virtual reality, intelligent building video surveillance and other fields. In this paper, we regard the "people " from the video image as the goal, using contour extraction and morphological expansion, finally, we can detect the number of people in the intelligent building by computer the area of the connected domain.Getting the original image is the prerequisite of number statistics, in this paper,we got the required image for post-processing by studying the imaging characteristics of different angles to determine the different placement of the camera.Image pre-processing of digital image processing is prerequisite for almost all of image processing, in this paper, we applied the image gray transform and image filtering. Improved the quality of the original image by histogram equalization and gray level histogram specification. Comparison of the mean filter, median filtering and Gaussian filtering method pros and cons, and choose a different noise depending on the original image for edge detection of the image late.Image edge detection has been the basic problem of pattern recognition, and also one of the most difficulties, this article compares the different effects of edge detection operator and a different scope, and ultimately selected the canny edge operator to extract obtained a better profile, to avoid the follow-up of complex processing.Finally, we solve the problem of discontinuity by expanding the edge of the image and filling the target image, connected domain area was calculated by setting the area threshold then, excluding the small image defects, and ultimately get the number of people. For the question of repeat count, this paper computed the area of connected domain under the threshold set, when the connected components is larger than a certain threshold, the goal counts for two people, finally, we solved this problem well.