Dissertation > Excellent graduate degree dissertation topics show

Algorithm of Small Target Detection in Strong Light Level Background

Author: PengZuo
Tutor: LiuZeJin；ZhangQiHeng
School: National University of Defense Science and Technology
Course: Optical Engineering
Keywords: Bright background Small target Low contrast image enhancement Noise, clutter suppression Segmentation Target detection
CLC: TP391.41
Type: Master's thesis
Year: 2007
Downloads: 166
Quote: 0
Read: Download Dissertation

Abstract

Low contrast, low SNR target detection technology is one of the key technologies of the photoelectric imaging detection system. This paper aims to solve the problem of small target detection in bright backdrop in accordance with the requirements of the tracking imaging detection system. The paper first analyzes the image and noise of small targets under the bright background clutter in the distribution of the characteristics of the airspace, and image filtering, enhancement, segmentation and testing stages were carried out corresponding algorithm. First, the use of the target and noise, clutter differences exist in the spatial form and related research using grayscale morphological bandpass and three in a row correlation filtering noise, clutter suppression method. The grayscale morphological bandpass filtering theory destination reserved a scale within the target signal and eliminate a large number of independent high luminance noise, as well as connectivity high low light noise, can inhibit a large area clutter; based on three in a row The filter in the introduction of further small target energy accumulation and noise, clutter suppression dual purpose. The experiments show that the proposed algorithm has better noise, clutter rejection performance. Second, through the imaging characteristics of the small target in the bright background analysis, research and the use of incomplete Beta function enhancements and piecewise linear enhancements to enhance small target image. Incomplete Beta function normalized advantages: select reasonable parameters α, β, converting the very steep curve in the vicinity of the need to stretch gradation segment, can be more greatly expand their dynamic range, so as to capture the low contrast The image of the target and background only a few critical difference grade. Regional processing based on engineering considerations, in this paper, a list of unknown origin simplified way to improve the real-time nature of the algorithm. Piecewise linear transform image enhancement and pattern recognition using minimum error method to do is transform the staging point to strike discriminant error probability minimum global optimal threshold, also reached better enhancement effect. Research and analysis of the main difficulties of the conventional segmentation algorithm in the presence of low contrast small target image segmentation of the pretreatment images taken two split way. Based on the threshold of a split, take full advantage of the small target and noise, clutter differences in regional connectivity, a region growing approach to image two split, proposed a simplified approach to overcome the traditional higher false alarm method to effectively extract the small target and enhance the robustness and the real-time nature of the algorithm. Tracking before detection due to traditional DBT detect method () can not be accumulated energy of the image in a very small target, the paper uses a the TBD detection methods (detection and tracking) dynamic planning theory, and based on the direction of constraints, detection pipeline continuous The three correlation filter and piping pipeline processing method simplifies algorithm, successful detection of small targets under the bright background. This thesis focuses on the difficulty in the small target detection technology under the bright background, combined with engineering practice, some highly targeted algorithm ideas and programs to improve the detection capabilities of current optical detection system provides a technical way.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile