Application In Transport Target Detection Based On The Dempster-Shafer Theory Of Fusion Algorithm

Traffic incident detection is one of the important researches of intelligent transportation system. The paper studies on the traffic incident detection fusion algorithm based on the D-S evidence theory.For the application of tunnel fire detection and road pedestrian detection, fusion algorithm based on the D-S evidence theory is studied in this paper, and corresponding improvement algorithms are put forward in this paper. The improvements include: combining credibility from the evidence and uncertainty to correct the weight values of evidence; Using the similarity and conflicts between the evidence to construct the K-L distance and get the weights of the evidence. Then fuse the weighted evidence. The improved algorithms are used in tunnel fire detection and pedestrian detection. First, extract the background from the video sequences using statistical histogram method, select the maximum pixel value as the pixel of background. Through subtraction method, the moving objects can be extracted from the background, while the binary processing and connected domain labeling can be executed afterwards to eliminate interference targets. Features in tunnel fire detection include: area growth characteristic, flashing characteristics, complexity of the shape and textures of smoke. Pedestrian detection uses aspect ratio, size and speed to detect pedestrians. Use mentioned features to identify and analyze the suspected areas in the videos. Simulatio n results of the detection can be generated by Monte Carlo simulation, Use MATLAB to fuse detection rate, false alarm rate. Preliminary experiments show the effectiveness of the algorithms.The results show that, compared with the single-source algorithm, fusion algorithm showing a better performance with a higher detection rate and a lower false alarm rate. The improved D-S evidence theory fusion algorithm has an obviously increase in the detection rate than the traditional D-S evidence theory algorithm with a lower uncertainty. The improved algorithm can be better in dealing with the high conflicts between two evidence, which can reduce the conflicts to a large extent. When adding a new evidence to fusion system, the detection rate can be further increased. The tests verifying that the fusion algorithm has better accuracy and efficiency.