FOD Detection Using a Multi-channel Information Fusion Method

Abstract

The Foreign Object Debris (FOD) on the airport runway threatens the safety of airport operations and is harmful to taking off and landing of aircraft. The traditional human eyes detection method has low detection efficiency and does not perform well in the detection of tiny objects. With the development of artificial intelligence and machine vision, detection systems using vehicle-borne cameras have emerged. This paper uses the 3D laser line scanning camera to obtain three-channel information, and proposes a FOD detection method based on multi-channel information fusion and Convolutional Neural Network (CNN). To improve detection accuracy, we propose two multi-channel information fusion methods to fuse three-channel information obtained by the 3D laser line scanning camera as the input of the network, and propose an improved YOLOv5 network which combines Backbone network and Convolutional Block Attention Model (CBAM) to improve feature expression capability. Our proposed methods are validated on our FOD dataset, which show that our proposed network has higher accuracy compared with the original YOLOv5 network, and our improved information fusion method which extracts image edges by Sobel and Laplacian operation is more effective than original information fusion method.