Detection and tracking of small targets in maritime infrared imagery

Infrared small and dim target detection is a key technology which is extensively used in military reconnaissance, navigation, security surveillance, missile guidance and other applications. But the detection of an object in maritime environment in infra-red images is a challenging task in computer...

Full description

Saved in:
Bibliographic Details
Main Author: Muruganandan, Raghavi
Other Authors: Deepu Rajan
Format: Final Year Project
Language:English
Published: 2016
Subjects:
Online Access:http://hdl.handle.net/10356/66732
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Infrared small and dim target detection is a key technology which is extensively used in military reconnaissance, navigation, security surveillance, missile guidance and other applications. But the detection of an object in maritime environment in infra-red images is a challenging task in computer vision and image processing due to the presence of random background noise, small and dim targets, lack of information regarding the shape or texture of the target, etc. However, with the help of a combination of various image processing techniques, it is possible to detect the small moving targets in maritime background. In this report, an algorithm is developed to detect distant targets (ships, boats) and track them in maritime infrared image sequences. As the small and dim targets are present at a very far distance from the infrared imaging system, they lie close to the sky-sea line. Prior to the identification of the sky-sea line, the image is converted to grayscale and a combination of normalized box filter and Canny edge detector is used to detect the edges in the image. The sky-sea line is then identified by finding the longest connected component/edge amongst the detected edges. As the target appears fused with the sky-sea line, a morphological erosion operation is performed to extract the target followed by clustering to group the remaining white pixels into clusters, after which a red box is drawn around the largest cluster (target). The algorithm performs fairly well in detecting and tracking the target when applied to two datasets, each consisting of 500 frames.