Research On Clutter Quantification And EO Imaging System Performance Prediction In Cluttered Background

The use of new materials and new technology has sharply improve the sensitivityand resolution of the electro-optical imaging system, meanwhile the background clutterbecomes the key bottleneck of predicting and evaluating the electro-optical imagingsystem performance. On the basis of different theories, this article proposed two classesof clutter metrics. Moreover, the target acquisition performance model suitable for theprediction and evaluation of modern electro-optic imaging system performance incluttered environment is presented by connecting the clutter metrics to the NVThermIPmodel statistically. The followings show the research work in details:1. With a deep research on the existing background clutter metrics and theconcept of clutter, several novel clutter metrics, such as relative complexity cluttermetric, structure similarity clutter metric, structure of edge background clutter metric,sparse-representation-based clutter metric and contrast-sensitivity-function-based cluttermetric, are presented based on background clutter characteristics. Further more, theinfluence of background clutter characteristics (like edge structure, sparse property andresolution) on target acquisition performance is discussed by analyzing the coherencebetween the clutter metrics and subjective experimental results of Search2database.2. Hidden-Markov-Model-based clutter metric is proposed by utilizing target andbackground clutter characteristics, when human visual perception is considered duringtarget acquisition process. Simulating the principle of recording a target in human brain,the target HMM is, firstly, obtained by training HMM with the target2D DiscreteCosine Transform coefficients. Secondly, the similarity between the target andbackground, serving as a novel clutter metric, can be calculated by decoding the optimalpath of target and background in the target HMM, the process of which accords wellwith the physiological mechanism of optimizing the search path of human visual system.Moreover, the experiment results show that proposed clutter metric is more consistentwith the subjective detection probability than its competitors.3. Unlike its ancestors modifying the cycle criterions with the clutter metrics, thisarticle constructs local target detection probability model and similarity target detectionprobability model for local target image and global background image, respectively.Then, the target acquisition performance model revised by the background clutter is gotby combining those two probability models. Compared to traditional performancemodels, the experiment results demonstrate that the revised model can more accuratelypredict and estimate the target acquisition performance of electro-optical imaging systems in real scenes.