The Vehicle Detection Based On The Saliency And LBP Method

The car as a means of transport, more and more widely into the people’s life with the development and improvement of people’s living standard. As the same time, traffic safety problems have cropped up more and more severe, in order to reduce the frequency of traffic incidents and facilitate people’s life, Intelligent vehicle technology is introduced. intelligent vehicle detection as an important part of driver assistance systems, its technical development has become even more perfect. Vehicle detection mainly by vehicle camera to monitor the front or rear of the vehicle environment, and process static image or video to comprehension immediate environment of the vehicle, including the vehicle size, color, position in the camera coordinate system, peripheral vehicle traffic environment, light environment and so on. Improve the accuracy and efficiency of the vehicle detection, reducing the error rate and the missing rate, it is important for late applications, such as accident monitoring, traffic monitoring, emergency intervention, collision detection, vehicle tracking and other applications.The vehicle detection algorithm is generally divided into two main parts, vehicle existence assumption and vehicle existence validation, the first phase to produce cars interested area, the second phase of interested area for classification and discrimination. This paper mainly for highway environment, according to the two stages of vehicle inspection and to implement the algorithm. This paper introduced the theory of visual saliency in the first stage, adjust the grayscale and saliency value as the pixel values of the weights and thresholds, that shadow block can still be retained and will be ruled out more interference factors, gets a better result than single processing grayscale, provide a more reliable vehicle for the following interested area.For the car shadow block, extract and merge car shadow line, to get preliminary interest areas, analyze the horizontal edge line of the vehicle accurately locate the vehicle area, combined with the symmetry of the vehicle and the vehicle information entropy to judge preliminary interested area. In vehicle existence validation phase using Adaboost classifier based on LBP feature vehicle identification for interested areas. In this paper, based on the significance and the car shadow block vehicle interested region extraction algorithm, comparison based on grayscale car shadow line for vehicle recognition and detection algorithm, the vehicle interested region extraction phase can reduce a lot of interferences, reduce the burden of the late recognition of operations such as, using the classifier can be further classification for interested areas, from the point of the experimental results, this algorithm can guarantee the car shadow line detection rate, reduces the error detection rate, and can satisfy the requirement of realtime.