Real Time Machine Learning Based Car Detection in Images With Fast Training

Abstract

Our primary interest is to build fast and reliable object recognizers in images based on small training sets. This is important in cases where the training set needs to be built mostly manually, as in the case that we studied, the recognition of the Honda Accord 2004 from rear views. We describe a novel variant of the AdaBoost based learning algorithm, which builds a strong classifier by incremental addition of weak classifiers (WCs) that minimize the combined error of the already selected WCs. Each WC is trained only once, and examples do not change their weights. We describe a set of appropriate feature types for the considered recognition problem, including a redness measure and dominant edge orientations. The existing edge orientation bin division was improved by shifting so that all horizontal (vertical, respectively) edges belong to the same bin. We propose to pre-eliminate features whose best threshold value is near the trivial position at the minimum or maximum of all threshold values. This is a novel method that has reduced the training set WC quantity to less than 10% of its original number, greatly speeding up training time, and showing no negative impact on the quality of the final classifier. We also modified the AdaBoost based learning machine. Our experiments indicate that the set of features used by Viola and Jones and others for face recognition was inefficient for our problem, recognizing cars accurately and in real time with fast training. Our training method has resulted in finding a very accurate classifier containing merely 30 WCs after about 1 h of training. Compared to existing literature, we have overall achieved the design of a real time object detection machine with the least number of examples, the least number of WCs, the fastest training time, and with competitive detection and false positive rates.