The Research On The Counting Algorithm Of The Traffic Flow Parameters Based On The Visual Tracking Technology

Along with the rapid development of the economy in our country, the progressively growing vehicle population leads to the degradation of the road traffic. As an important part of the ITS, the traffic management technique based on the video surveillance has got much more attention. The traffic flow parameters play an important role in the traffic management and the urban road constructions as well. The counting of the traffic flow parameters based on video surveillance can not only raise the intelligence level of the city’s traffic, but also reduce the investment of the unnecessary manpower and material resources. On the basis of visual tracking technologies, the thesis has focused on the transportation unity detection, features extraction, targets tracking, and the counting of the traffic flow parameters. The main works of the thesis are as follow:(1) In target detection, the thesis has studied the common algorithms of target detection for the rigid transportation entity, in which the vehicle are a representative. Because of the differences between the rigid and the non-rigid transportation entities in the deformation of the targets while moving, the thesis proposed an algorithm to do the detection of the pedestrians which combines the binary images and the corner detection. The algorithm works well in eliminating the errors led by the deformation of the pedestrian’s limbs while walking.(2) In features extraction of the vehicle, the thesis tends to remove the detection errors through calculating the proportion of the numbers of the pixels in the main contours and the noises. The body deformation of the pedestrian while walking will bring errors to the features detection. In order to solve this, the thesis proposed a feature-extraction algorithm based on the weights of the limbs. While extracting the features, since the front limbs, trunk and the back limbs have different impacts while moving, they have been entitled different weights. Then the feature point of pedestrian is calculated through weighted summation.(3) In targets tracking for the vehicle, in order to solve the problem that the similar background color interferes the targets tracking, the thesis choose the algorithm based on the combination of the Camshift and the Kalman filter, in which the former is to calculate the best position of the targets in the video surveillance and the letter is to provide the predicted positions for the targets. As for the pedestrian tracking, considering that its self-deformation will bring influence on the tracking operators, the thesis selected the multi-targets tracking algorithm based on the particles filter. The algorithm can complete the tracking through the reasonable distribution and the optimization of the particles. The algorithm to count traffic flow parameters based on the visual tracking technique shows a great advantage in eliminating the statistical error brought by the changing of moving targets states.(4) In traffic flow parameters counting, the common algorithms need preset parameters which makes the generalizability and the flexibility lower. To solve it, the thesis proposed a new algorithm on the basis of the computer vision and the thought that human use while counting. The algorithm needs no preset parameters. It completes the counting according to the comparison between the targets’ features in the two consecutive frames, and then achieves other traffic flow parameters. On the basis of the research results, a system applied to count the traffic flow parameters based on targets tracking has been designed and implied. The experiment result shows that the algorithm has a non-less-favorable function and a high generalizability.