Diksha L. Meshram;
In recent year terrorist attacks and different malicious acts usually target public areas and unguarded areas, like transport hubs, shopping malls, town centers, parks, and restaurants. To avoid this type of dangerous attacks that caused by terrorist and different crime related suspect a X-ray primarily based scanning system is accessible. This paper present a work on developing a system for automatic detection of threat items by using X-ray baggage security screening that facilitate us to reduces the risk of crime and terrorist attacks. This system created for characteristic threat objects that are photograph based on implicit shape models. We have a tendency to check the effectiveness of our technique for the detection of two different threat objects: i) razor blades; and ii) handguns.
Index Terms — X-ray image, threat objects, razor blades, handguns.
X-ray imaging provides a significant tool in checked baggage inspection, with various sensitive techniques being delivered in touch in determining the form, and density of things within luggage also as other material dependent parameters the potential of finding these objects effectively may be a crucial concern for national security. Currently, the detection of prohibited items relies on Transportation Security Officers (TSOs) to visually detect this stuff from displayed image scans this is often challenging for several reasons. Firstly, the set of threat objects that security officers must identify is kind of diverse, sharp instruments; firearms; blunt weapons for security concerns. Second, the majority of scans are benign, yet TSOs must remain alert for long periods of some time. Third, because X-ray scans are transmission images, the contents of a bag appear stacked on top of each other into one, often cluttered scene, which can render identification of individual items difficult the prevailing X-Ray machines installed at check in area to screen registered baggage are in stand-alone mode and unable of detecting 100% explosives. Stand-alone machines utilize space at waiting area and also cause big queues, which disrupt passenger flow during peak hours. In addition an object inside a bag could also be in any position, it’s going to be rotated so an algorithm is rotational, translational invariant should be used for providing accurate results. usually threat objects are covered by another objects in the bags it become harder to detect it (by the effect of superposition). The passenger’s baggage may contain threat items like razor blades, handgun, bomb, etc which must be detected efficiently therefore the human operators must be assisted by a weapon detection system. The detection of threat objects relies on Transportation Security Officers (TSOs) to visually pick these items from displayed image scans this can be challenging for several reasons. First, the set of prohibited items that TSOs must identify is kind of diverse: sharp instruments;
ﬁrearms; blunt weapons. Second, the bulk of scans are benign, yet TSOs must remain alert for long periods of your time. Third, because X-ray scans are transmission images, the contents of a bag appear stacked on top of every other into one, often cluttered scene, which may render identiﬁcation of individual items difficult. For the aforementioned reasons, an automatic threat detection algorithm to assist human operators in locating threat objects would be useful for the TSOs, especially if it can be readily integrated into the prevailing of deployed scanners. Object detection has long been considered a challenging task for computers, but advances in deep learning in recent years have resulted in enormous progress. Speciﬁcally, Convolutional Neural Networks (CNNs) have proven extremely useful at extracting learned features for a large kind of computer vision tasks, including object detection. As a result the TSOs is fascinated by assessing the feasibility of deploying algorithms which will automatically highlight objects of interest to TSOs. Most deep learning methods require an outsized training dataset of labeled examples to attain good performance for object detection, this implies data comprising both images and bounding boxes with class labels. In order to produce appropriate security, a far more sophisticated, reliable, and fast screening technique is required for passenger identification and baggage examination. Automatic threat detection is a crucial application in x-ray scene analysis. Understanding x-ray images may be a challenging task in computer vision and an automatic system should be developed that consumes less time for processing and performs accurately with reduced false positive results.
II. RESEARCH METHODOLOGY
For the detection of threat items, X-ray imaging systems and MMW (Millimeter wave imaging) are used and x-ray imaging system is widely used for carry-on bags. The techniques used for analyzing these x-ray images are pseudo-coloring and segmentation based techniques. the Pseudo-coloring is one of the process into which the threat objects inside the bags are provide various colors based on their material type. In segmentation based methods, the x-ray images are segmented to extract the features of the objects of interest. Using these methods, satisfactory results are produced and assisted human operators for detecting the threat items. X-ray photons however, penetrate most of the materials. As a result, all threat items along an x-ray path attenuate the x-ray and contribute to the intensity of final measured. In the x-ray community, a common way of disambiguating threat objects is through CT reconstruction. This
is typically obtained through filtered back projection algorithm. Whereas different types of X-ray technologies are based on automatic systems are exist for threats detection, only few of them systems makes use of the well-established pattern recognition and machine learning techniques. Recently X-ray imaging systems at airports use dual-energy analysis to estimate the atomic numbers of materials within the passenger baggage. This method obtains a measure of the density and thickness of material.
III. PROPOSED SYSTEM
The matching between two images of scenes or detection of any object among different objects is part of computer vision. To perform matching between two images the point correspondences is needed. This task of matching is divided into three parts, firstly detection or identification of interest points, secondly description of that interest point and third is to find correspondences between two images. Detector performs the task of identification of interest point. Interest points are expressive in texture which help us to modify any desired object or scene among different undesired objects. Interest point could be the point at which the direction of the boundary of object changes abruptly or it can be the intersection point between two or more edge segments. These are different from edges as edge is in particular a line segment on the boundary where two faces meet or it is often called as side. Finally, descriptor vectors (feature vectors) are then matched between two images. The distance between these two vectors is calculated first for ex. Euclidean distance and based on this distance matching is carried out. To increase the speed of interest point matching the dimension of feature descriptor should be low. However, feature descriptor with low dimension is less distinctive than high dimensional descriptors.
IV. EXPERIMENTAL RESULT
The figure 3 shows the object database. The input object database is created. This dataset feature is calculated by using SURF algorithm. The key points drawn by the SURF algorithm are shown in figure 4. The figure 5 shows the detected output of threat object blade and gun.
V. CLASSIFICATION RESULTS OF DIFFERENT THREAT
Security is important concern regarding to national and international security. The x-ray image based threat detection provides a significant tool in checked baggage inspection. The various threat images are required which can be treated for learning of computer. The detection of such images in image processing is not an easy task. The various work need to be required to make the system work accurately and detection of threat can be detected very easily and automatically.
- Vladimir Riffo and Domingo Mery, “Automated Detection of Threat Objects Using Adapted Implicit Shape Model”. IEEE 2016.
- Sebastian Flores1 • Domingo Mery2, “Threat Objects Detection in X-Ray Images Using an Active Vision Approach_springet” 2017.
- Trevor Morris,Tiffany Chien, Eric Goodman, “Convolutional Neural Networks for Automatic Threat Detection in Security X-ray Images” IEEE 2018
- Thomas W. Rogers, Nicolas Jaccard, James Ollier, Lewis D. Griffin, “Threat Image Projection (TIP) into X-ray images of cargo containers for training humans and machines.”IEEE 2016
- Samet Akc¸ay1, Mikolaj E. Kundegorski1, Michael Devereux2, Toby P. Breckon1,“TRANSFER LEARNING USING CONVOLUTIONAL NEURAL NETWORKS FOR OBJECT CLASSIFICATION WITH IN X-RAY BAGGAGE SECURITY IMAGERY”IEEE 2016
- Trevor Morris, Tiffany Chien ,Eric Goodman S, “Convolutional Neural Networks for Automatic Threat Detection in Security X-ray Image ”, 2018 17th IEEE International Conference on Machine Learning and Applications.