FOD detection using Intelligent Computer Vision Techniques

Project Title

FOD detection using Intelligent Computer Vision Techniques

Project Area of Specialization Artificial IntelligenceProject Summary

Foreign Object Debris (FOD) is a substance, debris or article that is found typically on the runways and cause severe damage to the aircrafts. This paper describes a video-based system used for FOD detection application. Research is going on throughout the world to develop an automated FOD detection system. Different companies are using different sensors for this purpose including millimeter wave radars, thermal sensors as well as electro-optical sensors. There are number of techniques used for object detection. One of the techniques is deep learning which is one of  the prime area of focus of research these days because of its performance as compared to the traditional image classifiers which were built on handcrafted features. An object detection algorithm based on the CNN will be developed which is capable of classifying and detecting the object with great accuracy irrespective of object orientation and poor lighting conditions in the image.

Project Objectives

Foreign Object Debris (FOD) found on runway is a safety hazard for aircraft and causes a huge amount of damage if ingested by the aircraft. Different schemes and sensors are there which have been utilized for FOD detection. In Pakistan, daily in the morning FOD walks are conducted at the runway for FOD detection which consumes a lot of man hours. Classical computer vision techniques are there which can help to detect FOD but these classical computer vision techniques use hand coded features which are not robust to different lighting conditions and would fail when exposed to different object orientation. To address these issues in this project, intelligent computer vision techniques will be used to classify as well as detect the FOD.

Project Implementation Method

First of all, getting familiar with different neural networks for object detection, once done with selection of suitable method, algorithm will be developed. Then a Deep Convolutional Neural Network (CNN) model will be trained to detect objects from a video captured by a high-resolution camera. The network will be learning over thousands of training images collected from both online databases and data that have been collected myself.

Benefits of the Project

1. This project will introduce an automated system for FOD detection which will reduce the man power that is consumed at the airports for FOD removel
2. The project will improve air safety as FOD is one of the major source of accidents on the runway

Technical Details of Final Deliverable

1. Selection of suitable computer vision technique for FOD Detection
2. Development of FOD dataset for training
3. Training of selected computer vision technique over collected dataset
4. FOD detection, testing and accuracy optimization over FOD images.
5. FOD detection from a video feed

Final Deliverable of the Project Software SystemCore Industry OthersOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Quality EducationRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	53000
Nvidia Jetson Nano	Equipment	1	35000	35000
Raspberry PI 4	Equipment	1	18000	18000