Research On Clutter Metric Scale Based On Machine Vision

With the continuous development of the Automatic Target Recognition(ATR)system,background clutter has gradually become an important factor affecting its performance.Background clutter has different effects on different ATR algorithms.This paper studies the effects of background clutter on image segmentation and target recognition algorithms.,a detailed introduction to some commonly used algorithms.On the basis of observing the basic principle of clutter metric,a reasonable clutter metric is proposed to quantify the clutter strength,and the relationship between background clutter strength and algorithm performance is preliminarily explored.In this paper,the clutter characteristics are studied.The concept and definition of background clutter are briefly introduced.The common clutter metrics are briefly described,and their advantages and disadvantages are analyzed.Their commonalities are summarized.Based on the metrics of human visual clutter,based on the characteristics of ATR algorithm,the clutter metrics suitable for these algorithms are proposed.According to the range of clutter metrics,the global clutter metric and local clutter metric are studied separately.The local clutter metric part introduces the image segmentation algorithm in detail and uses segmentation precision to characterize the performance of the algorithm.The local clutter metric is based on the LBP feature,which is a common image texture feature.The image texture features and LBP features are introduced in detail,and their important research significance is analyzed.The simulation proves that there is a monotonic relationship between the intensity value of the local clutter and the segmentation accuracy of the segmentation algorithm.The global clutter metric is based on the HOG image feature.The feature is introduced in detail in this paper.The implementation process of the metric scale is described.The relationship between the global clutter strength and the false alarm rate is verified by simulation.In this paper,the principle of small target detection is studied in detail.The difference between the detection principle of infrared small target detection algorithm and common target detection is analyzed.It is shown that the commonly used metrics cannot be adapted to small target detection.A new clutter metric method is introduced for the characteristics of small target detection algorithm.Finally,the effectiveness of this method is verified by simulation.