By S. Asnani, A. Ahmed, A. A. Manjotho
Keywords: Cascade Classifier, HOG features ,Object Detection, Security System, and Unconcealed Gun Detection.
Banks play the most important role in currency circulations almost everywhere in the world. Due to this reason, banks have become the target places of criminals. This paper proposes a solution for providing safe and healthy work environment to the personnel. The proposed security system is based on Digital Image Processing techniques. The idea is to detect the visual unconcealed weapons whenever they appear inside banks. The major problem in training any classifier is to select the appropriate features from the images so that the classifier can work with greater accuracy and efficiency. Through various research work and experiments, we have found that Histogram of Oriented Gradients (HOG) features have proved to be a robust feature for detecting many different objects with high detection accuracy. HOG was introduced by Dalal for detecting humans, but it can also be used for detecting any other type of object equally well. A boosted cascade classifier has been used for training purpose using the HOG features. A large collection of training dataset is used and the final detector is tested on larger testing data.