Vehicle Detection and Recognition by Correlation Methods (1190 words)

Vehicle Detection in Static Images using Matlab. The "Neural Networks for Image/Pattern Recognition" is vast and we should limit ourselves with one type of object recognition which is Vehicle.

In this Project we have to :

1. Provide the references used for building the project or concept used in the project.
2. Count the numbers Vehicles in the Static color or B/W Image.
3. How to distinguish what type of vehicle is in the Static color or B/W Image (car, truck, bus or motorcycle)

This material may consist of step-by-step explanations on how to solve a problem or examples of proper writing, including the use of citations, references, bibliographies, and formatting. This material is made available for the sole purpose of studying and learning - misuse is strictly forbidden.

Abstract

The purpose of this project is to implement a correlation method for vehicle detection and vehicle recognition in BW images. The initial goal was to use Hopfield neural network approach. We found this method to be unstable with increasing size of the training database, and the resulting classification of no particular value.

We found the correlation method to be quite reliable in the classification of cars vs. non cars, and present the main result in the form of the confusion matrix for the detection algorithm.

Introduction

The automated detection of objects in images and video has been a very active research area for the last 30-40 years. The detection task is important for mapping, navigation, surveillance and monitoring, and tracking, and is an essential task in making sense of the environment. A multitude of techniques has been applied for use on a variety of platforms. A common approach is to segment a scene into specific objects, for instance using the SIFT algorithm, on the basis of context organize the objects detected and classify these objects into composite objects as a means for classification, for instance by the development of suitable matched filters. An important component of the detection algorithms is a correlation between the objects detected and those in a database of recognized objects....