Description

The video based vehicle detection method to capable of continual operation under various environmental and illumination conditions. And they identifying features that can adapt to various conditions is crucial. For this paper proposes a robust vehicle detection method for identifying horizontal edges by using a Sobel filter to achieve low complexity. Based on the orientation of the gradient feature of a vehicle can be extracted accurately and quickly. They are mostly symmetrical features are critical features, and a histogram of oriented gradients was employed to reduce the detection error rate. It is more advanced system are used to the suitable for integration with advanced driver assistance systems and to show the forward vehicle horizontal & vertical condition easily. Such that cases for the intensity to illumination of situation for Tunnel, Rainy, Sunny, Freeway for this timing situation. The detection of the vehicles based process has major applications in the development of the automated surveillance system. We propose an object based segmentation method that locates and identifies the objects in the input video. The object approaches were robust to illumination variations. The proposed vehicle detection system uses the gradients and the selected ROI for the identification of the vehicle. The input video is initially converted into frames the frames were then preprocessed using median filter. The edge regions were detected from the preprocessed frames using Sobel operator. The detected edge pixels were then used for the calculation of the gradient information. The image pixels present with the detected ROI were chosen. The ROI is selected manually by eliminating the road region in the frame. The pixel similarity between the ROI and the detected edges gives the gradient information of the frame. The obtained gradient informations has a combination of the detected object and other regions in the image. The gradient informations removes the regions other than road in the video frames. The obtained gradient is filtered using morphological operations. The morphological operation such as dilation is employed inorder to get the objects in a better manner. Then by measuring the area of the objects in the overall frames the other regions in the images were removed. The detected objects were then marked inorder to identify them separately. The performance of the process is measured by measuring the accuracy and error rate of the detected objects in each frame. The accuracy and the error rate measured shows that the proposed method is capable of the identification of the vehicles in the video in an effective manner. The overall accuracy of the process ranges from 85 to 87%.