Detection Method Of Forest Fire False Alarm Source With Infrared Imaging Feature Modeling

Space infrared monitoring system plays an important role in early warning and missile interception,and forest fire is a common false alarm source in military space infrared monitoring system.Through the study of characteristic modeling and detection method of forest fire in infrared remote sensing image,the disturbance of false alarm source of forest fire is removed in the image first,so as to effectively assist the detection of small infrared targets such as missiles,and also help to confirm the scope and scale of fires.This thesis focuses on the detection of forest fire area in infrared remote sensing images,studies the characteristics of forest fire infrared imaging,extracts and selects the image features,and proposes a feature selection method based on distributed overlap probability.On this basis,two detection schemes of suspected forest fire in infrared images are designed,one is based on the image significance feature,the other is based on the edge features.In addition,in the aspect of region segmentation,an adaptive threshold segmentation method combined with regional characteristics is proposed,which can make the segmented target more complete.Finally,based on the extracted statistical features of forest fire area and background area,the support vector machine classification model is established to identify the suspected area and eliminate the false alarm area.Experiments show that the whole scheme can accurately detect the forest fire area in the infrared remote sensing image.The main research contents are as follows:(1)In this thesis,the imaging characteristics of forest fire in infrared remote sensing images are studied,and the appropriate image preprocessing method is selected according to the characteristic analysis.(2)This thesis studies the scheme of image feature modeling,extracted the image edge features and saliency features,established the sample set of forest fire area and background area,extracted the statistical features of the sample set,and then proposed a feature selection method based on the distribution overlap probability.The method is used to eliminate invalid features and establish a feature space.(3)The detection method of forest fire area in infrared remote sensing image is studied.Two detection schemes are proposed,one is based on the significance feature and the other is based on the edge feature.In the detection scheme,an adaptive threshold segmentation method combined with regional characteristics is proposed to achieve more accurate segmentation of suspected area.(4)The pattern recognition theory is studied.By using the statistical features of the sample set,the classifier of support vector machine(SVM)is trained,which can distinguish the suspected fire area and eliminate false alarm area effectively.