The Study Of Vehicles Detection Based On Part Model In Natural Scene And Its Parallel Methods

A variety of methods have been adopted to collect numerous pictures about vehicles in different industries, these pictures obtained most in natural environment in an automatic or semi-automatic way share some common c haracteristics, for instance, the collected scene is not fixed, the picture qualit y is not stable, the time period of the collected scene is in large quantity a nd well-distributed. To determine the relative position of vehicle in the pict ure efficiently and accurately is necessary to increase the availability of the collected image information. Moreover, nowadays most of the vehicle detect ion methods have strict criteria for factors like photo shooting angle, light an d shade conditions, which can only be achieved these in particular condition.Therefore, research on how to detect vehicle under natural scenes efficiently and accurately has both important theory and practice significance.A variety of methods have been adopted to collect numerous pictures about vehicles in different industries, these pictures obtained most in natural environment in an automatic or semi-automatic way share some common characteristics, for instance, the collected scene is not fixed, the picture quality is not stable, the time period of the collected scene is in large quantity and well-distributed. To determine the relative position of vehicle in the picture efficiently and accurately is necessary to increase the availability of the collected image information.Moreover,nowadays most of the vehicle detection methods have strict criteria for factors like photo shooting angle, light and shade conditions, which can only be achieved these in particular condition. Therefore, research on how to detect vehicle under natural scenes efficiently and accurately has both important theory and practice significance.With the widespread use of the GPU with its increasing performance, the GPU parallel computing technology is becoming more and more mature. Furthermore, the characteristics of GPU have determined that it has much more advantages than that of CPU in image processing. Therefore, this paper adopts the method of parallel computing, reducing the running time of the algorithm efficiently. In this paper, by using support vector machine to solve the vehicle detection under natural environment, the research results show that even when the scene is not fixed at all, the detection results can be absolutely good without artificial design feature.