Smart ATM With Tracking of Criminals Using Novel Di-Pattern and C-LDP (Combined Local Directional Pattern)

Introduction

Automated Teller Machines (ATM) offers much convenience to everyone in life due to their easy and readily available cash. Frauds related to the ATM are increasing day by day which is a serious issue. Some of the ATM issues are shown in Figure 1. ATM is equipped with surveillance monitor; criminals usually attack the customers and try to steal their wealth by occluding their faces. The thefts on ATMs are steadily rising and this is a serious problem for law enforcement and banking sectors. This paper mainly focusing on protection of customers inside the ATM using CCTV security cameras and emergency sirens.

Figure 1.

Different attacks in ATM

In different security systems enabled in banks and ATM’s, the automatic face recognition plays a major role in video surveillance. The alternate methods like finger print identification have major drawbacks like the need of co-operation of the suspect. The face recognition methods are overcoming those drawbacks and are more cost-effective. The face recognition applications play a major role in law enforcement. In most of the cases the photo database of the criminals maintained by the police are enough. But in certain situations when the photos are not available the sketch drawn through eye-witnesses come to a rescue. This work proposes an automatic detection of the criminals through such sketches with high accuracy (Benson & Perrett, 1991). Caricature is a special drawing of human faces which has the needed details of a human face. They represent the essential information. These caricatures can be used to recognize a human being. Bruce et al. (1992) have explained that the cartoons with shading and pigmentation details can also be well recognized by human beings. The computers can recognize the human beings using these sketches and cartoons at a better accuracy than the human beings. In most of the existing methods the photos are converted to sketch before recognition (Tang & Wang, 2002; Wang & Tang, 2009). Then, a patch-by-patch comparison is made. In the proposed work there is no separate conversion of photo to sketch representation and no separate patch-by-patch comparison. A novel C-LDP algorithm is devised that converts the photo and sketch to a code image of less variation. The proposed algorithm also creates a feature vector that has a patch information for better accuracy. The use of ELM classifier enables faster and accurate classification.

The paper is structured as follows: Section 2 presents the concepts of the proposed work. The experimental results are given in Section 3. Finally, conclusion and future work are presented in Section 4.

Proposed Work

The proposed system first classifies whether it’s a weapon or not and the based on criminal database facial identification is done. If both are matched, it gives the alert to the nearby policy stations.

• 1.

  Weapon detection

• 2.

  Facial identification