People Counting Method Based On Optical Flow In Video Image Sequences

People counting is an important research content in the Computer intelligent videosurveillance. People counting system commonly using object detection and trackingmethod. These systems analyze the trajectories of moving objects in the scenecorresponding mostly to individuals. Optical flow provides a fresh idea for us todesign a people counting system. Using optical flow to estimate the direction of themoving pedestrian, it doesn’t need to consider the trajectory of the moving pedestrian.For this reason, the people counting system became more simple and efficient. Thepaper made the following works:1. In the paper, an approximate size of pedestrian in the vertical view is proposedas a feature. People counting method based on optical flow and pedestrian Size areintroduced in detail. In this method, a virtual detection line is settled and optical flowalgorithm is used to calculate the optical flow vector of pixels on the line. An onlineblob detection algorithm is used after the optical flow vectors of the pixels on thecounting line have extracted moving blob. The method constructs the approximatesize of pedestrian in the vertical view to estimate the number of people in the movingblob, and count the people who passed through the detection line.2.In the paper, the color information has combined with the Lucas-Kanade opticalflow algorithm. Using five color models which are RGB,YUV,HSI,HSV and Lab tobuilt five color optical flow algorithms. Then, the six optical flow algorithms likeLucas-kanade and other color optical flow algorithms will estimate the optical flow inthe counting people method which has been mentioned.3.In the paper, the confusion matrix where tp(true positives), fn(false negatives)and fp(false positive) are used to estimate the precision and recall. Then F will bemeasured by the precision and recall, which is used to evaluate the people countingsystem. It is a reliable Evaluation method. And this paper will contrast theaccuracys,which are used in the people counting method with using Lucas-Kanadeoptical flow algorithm and other five color optical flow algorithms. Experimental results show that this method could reduce interference which iscaused by shadows and image preprocessing, and accurately count the number ofpeople passing through the entrance. Among the several optical flow algorithms, Laboptical flow algorithm has the highest accuracy in the people counting method.