Development of Airline Baggage Scanning System using Deep Learning

Passenger and carry-on luggage screening is one of the key front line security measure that helps ensure the security of international civil aviation .Therefore safety of the airport is one of the biggest priority for any nation. Transportation Security Administration (TSA) screeners at airports che

Project Title

Development of Airline Baggage Scanning System using Deep Learning

Project Area of Specialization

Artificial Intelligence

Project Summary

Passenger and carry-on luggage screening is one of the key front line security measure that helps ensure the security of international civil aviation .Therefore safety of the airport is one of the biggest priority for any nation. Transportation Security Administration (TSA) screeners at airports check passengers and personal items to detect dangerous items such as weapons, chemicals and liquids that are not allowed as carry-on items. But due to psychology of human brain a person cannot maintain focus on his work for more than 50 minutes and this  factor can enhance the probability to miss illegal items during screening and put the security of airport at stake. This project is aimed to provide artificial intelligence based screening system that will be trained and used to detect unwanted items in passenger’s baggage. This project will substantially reduce the probability of missing  unwanted items due to human  psychological factors.

Artificial Intelligence can simplify the process of baggage screening. Airport screener will create a 3D image of what's inside each bag. All the different materials will be represented with different pseudo-colours. Metal, for example, will tend to come out blue; organic material tends to come out orange; lighter metals tend to come out green. That can be interpreted automatically by deep learning which can detect this bag has maybe an illegal material. In an airport context, the machine could set off an alarm when it recognized a dangerous material. It could work in conjunction with a traditional 3D X-ray image, to show security workers where in the luggage the threat is located.

Currently, screening equipment such as metal detectors, millimeter wave machines, backscatter x-ray and cabinet x-ray machines are being used manually by civil aviation security. But due to human error and human brain phycology, probability of error is very large. Therefore Artificial intelligence based this model will significantly reduce chances of human error and strengthen security system.

Project Objectives

Identification of firearms, weapons, incendiary or explosive devices or other dangerous  devices by metal detectors ,X-ray units is main purpose. When items Of cabin baggage are screened by conventional X-ray, every images of every part of the item being screened are to be displayed for no more 5 seconds after which baggage should be rejected if it cannot be cleared by the screener. Baggage AI will function as a ‘second pair of eyes’ for the security personnel assisting them in bolstering security by detecting and effectively removing the minutest of threats. However Regulatory requirements for security screening at airports vary tremendously from State to State. Moreover, different threat levels will force some States to apply stricter requirements on a permanent basis. It should be clear that any increased threat should be met with the addition or substitution of additional security measures.

Project Implementation Method

The method that we are going to implement in our project is a well optimized algorithm of object detection which is CNN (Convolutional Neural Network). Neural Networks form the base of deep learning which are subfield of machine learning where the algorithms are inspired by the structure of human brain. Neural networks take in data train themselves to recognize the patterns in this data and then predict the output for a new set of data. Convolutional Neural network also known as CNN or ConvNet is an artificial neural network that is so far been most popular used for analyzing images, image analysis is the most wide spread use of CNN's. Most generally a CNN is an artificial neural network that has some type of specialization for being able to pick out and detect patterns and make sense of them, this pattern detection is what makes CNN so useful for image analysis. A CNN has hidden layers which are known as convolutional layers. A convolutional layer receives input then transforms the input and then outputs the transform input to the next layer with a convolutional layer this transformation is a convolution operation. For example, an image is taken as input. Then it is divided into regions. Each region is considered as a separate image and all of these regions are passed to the CNN, and it is then sent through various convolutional layers which classifies them into various classes. CNN extracts feature of each region and uses SVMs (Support Vector Machine) to divide them into classes. SVM is a supervised Machine Learning algorithm which is used for classification of objects. When each region is divided into its corresponding class, then every region is combined to get the final output image with the detected objects. We can use this technique to detect various objects in an image. One problem with Simple CNN algorithm is that it requires a large number of regions which require a large amount of computational time. There are advanced CNN algorithms which solves such problems like RCNN (Region-based CNN), Fast RCNN, Faster RCNN and YOLO. RCNN algorithm makes a lot of boxes in the image and checks if any of these boxes contain an image. RCNN uses selective search to extract these boxes from an image. Fast RCNN algorithm uses a single model which extracts features from the regions, these regions are divided into different classes, and the boundary boxes are returned for the identified classes simultaneously. Faster RCNN is an enhanced version of Fast RCNN. The major difference between them is that Fast RCNN uses selective search for generating Regions of Interest, while Faster RCNN uses RPN (Region Proposal Network). RPN takes image feature maps as an input and generates a set of object proposals, each with an accuracy score as output. Faster RCNN is less time consuming compared to the other algorithms.

Benefits of the Project

Over the years, the industry has come to recognise that an X-ray machine is only as effective as its operator. But humans are emotional creatures and far from perfect, so when the stakes are high it makes sense to invest in advanced technology to mitigate human error and enhance operations.Airlines companies discusses the role artificial intelligence might have on aviation security processes in the future. Baggage screening is tedious but important task that needs to be done at the airport.However, AI has simplified the process of baggage screening.AI technology developed to screen baggage for multiple passenger lanes.Such tools will not only automate the process of baggage screening but also help authorities detect illegal items effectively.World's leading airliner service providers are now using AI tools and technologies to deliver a more personalized travelling experience to their customers.Crucially,AI systems improve as more and more information is fed into them. In the case of airport security,machine learning can be used to analyse data and identify threats faster than human could. Machine learning techniques are used to automatically analyse data for threats,including explosives and firearms,while ignoring non-dangerous items-for example keys and belt buckles-users may be carring. Backscatter X-rays are designed to show hidden objects beneath clothing or luggage. They can detect both metallic and non-metallic objects, ranging from guns to foods and plastics. The computer vision system analyzes x-ray images by their shapes automatically gives alert when it 'sees' an illegal item. This leads to the reduction of labor-intensive screening. The system elliminates the need of tedious and time-consuming manual baggage inspection. This doesn't only speeds up the process but also serves as a security assurance to passengers.Artificial intelligence is poised to take the lead in enhancing aviation security beyond the limitations posed by human errors.X’ray  baggage security screening is widely used to maintain aviation and transport security and poses a signi?cant image-based screening task for human operators reviewing compact, cluttered and highly varying baggage contents with in limited time-scales.artificial intelligence (AI) emerged as a boon for aircraft companies. Now with the implementation of AI, aircraft companies can predict potential maintenance failures on aircraft before any mishaps occur.

Technical Details of Final Deliverable

In our project we are going to use an optimized algorithm for object detection which is CNN (Convolutional Neural Network). The CNN algorithm can be implemented by using Python’s tensorflow and PyTorch libraries. Tensorflow is an open source library created by a team of Google and it is used for Numerical computation for machine learning algorithms and deep Neural Networks. It is also used for image classification and object detection. PyTorch is also an open source library that works like tensorflow but its advantage is that that it can perform numerical computation on dynamic graphs used in deep learning neural networks. Both can be used for the implementation of object detection algorithms. There is a difference between image classification and object detection. An object classification model is used to classify an object when there is a single object in the image.  While an object detection model can identify which of a known set of objects might be present and provide information about their positions within the image. This model needs to be trained first with a set of data and then it identifies images with an accuracy score. The COCO (Common Objects in Context) dataset will be used to train the object detection model. COCO is a large-scale object detection, segmentation, and captioning dataset which contains images and bounding boxes. We will use a model which is trained on COCO Dataset. Then a graph will be generated by tensorflow of that model. All images will be converted into arrays for further processing. This model runs the inference for a single image, where it detects the objects, make boxes and provide the class and the class score of that particular object. Similarly inference will be run on each image and same procedure will be repeated again. Some images from the dataset will be used to train our model and once it is trained the remaining images will be used for testing of the model. Like this Live Object Detection can be performed using Python’s OpenCV (Open source computer vision) library. OpenCV is a library of programming functions mainly aimed at real-time computer vision. Live object detection can be used to detect objects in real time, like the video generated by X-Ray baggage screening machine at the airport. Live object detection uses image processing and computer vision which requires high end GPUs (Graphics Processing Unit) which are expensive.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Education

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Industry, Innovation and Infrastructure, Sustainable Cities and Communities

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com