An introduction to support Vector Machines: and other kernel-based learning methods
An introduction to support Vector Machines: and other kernel-based learning methods
On the algorithmic implementation of multiclass kernel-based vector machines
The Journal of Machine Learning Research
On-Road Vehicle Detection: A Review
IEEE Transactions on Pattern Analysis and Machine Intelligence
Detection and classification of vehicles
IEEE Transactions on Intelligent Transportation Systems
Hi-index | 0.00 |
The vehicle area segmentation is important for the various applications including ITS (Intelligent Transportation System). We present a novel approach for segmenting a vehicle area from still images of vehicles on the asphalt paved road captured from outdoor CCD cameras. Our algorithm classifies the partitioned grid areas in the input vehicle image into road or vehicle classes. Texture features are used for representing each class, and we use SVM (Support Vector Machine) method for the classification. Our preprocessing process partitions given sample images into a set of grids, and classifies each grid area into two classes: i) road class, and ii) vehicle (non- road) class. We use GLCM technique to extract the feature values for each class, and sample classes are trained by using the SVM. The SVM constructed in preprocessing step is applied for each given input image to decide whether the grid in the image belongs to the road area or not. After marking the grids as road or vehicle classes, we find the optimal rectangular grid area containing the vehicle. The optimal area is found by using a dynamic programming technique. Our method efficiently achieves high reliability against noises, shadows, illumination changes, and camera tremors. We experimented on various vehicle image set, where the images in each set are captured in different road environment. For the largest set, by using 50 sample images, where each image with 1280 ×960 resolution or 13 ×12 grid areas, our algorithm shows 94.31% of successful vehicle segmentation from 211 images with various kinds of shadows and illumination changes.