Regional Traffic Status Detection By Combining Haar-like Features With Spatial-temporal Information

With the high traffic demand and the high density of urban population,the problem of urban congestion is increasingly serious.At present,from the supply side and the demand side,the government has implemented the policy of increasing the construction of transportation infrastructure and restricting vehicle license plates and traffic.However,it can’t fundamentally solve the structural problem of traffic congestion.more scientific,efficient and intelligent methods are needed to detect and manage the traffic status.At present,intelligent traffic system and intelligent traffic are constantly developing and improving.video-based traffic detection technology has become a research hotspot due to its characteristics of visualization,non-contact,economy and multi-detection.This paper proposes a regional traffic status detection algorithm that integrates Haar-like features and spatial-temporal information.The main contents are as follows:First,the basic concept of traffic state is introduced,and a new definition of local area of traffic state is put forward.At present,the method takes the traffic condition of the road as a whole to distinguish.However,urban traffic is complex and changeable.In the same road,local areas with different lanes and different distances may also have different traffic conditions.This paper gives the definition of the traffic status region,further refines the traffic status detection,and meets the accuracy requirements of the urban complex scene.Second,the three traffic status detection technologies are introduced and their advantages and disadvantages are analyzed.The Ada Boost detection method based on Haar-like can directly detect vehicles to obtain the traffic status on the road,but the problem is that the remote small target and the blocked target cannot be recognized.Canny edge method and frame difference method can respectively obtain the spatial existence and motion of objects on the road,and are often combined to detect the road traffic status.But Canny edge method has problems with noise and non-target recognition.Frame difference method has "double shadow" problem and "void" problem.Third,the traffic status detection algorithm in this paper is proposed,which divides the road into lanes and local areas,and combines the advantages of the above three algorithms to solve their defects by learn from others’ s strong points to offset one’s weakness to detect the traffic status areas.First,Ada Boost detection method based on Haar-like features,Canny edge method and frame difference method respectively detect the road image to obtain vehicle,spatial information and temporal information.Second,through the algorithm in this paper,the obtained detection results are converted into binary detection signals with outputs of "0" and "1".Third,through the analysis of road traffic status,a signal discrimination method is proposed,which comprehensively judges the detection signals and extracts the regions of different traffic status on the road.The failure problem of detection signal caused by the defect of the above algorithm is also supplemented and corrected by integrating other detection signals to divide the accurate traffic status area on the road.Finally,in order to prove that the algorithm in this paper is helpful to detect traffic parameters,the vehicle queue length is estimated on this basis.Fourthly,the algorithm in this paper is programmed.The training process of Ada Boost classifier based on Haar-like feature is introduced in detail.The algorithm in this paper is compared with other two algorithms in complex traffic scenarios.The results show that the algorithm in this paper has better performance in the urban complex traffic scene,satisfies the accuracy and stability requirements of traffic state detection,and refines the traffic state detection results on the road.