Model Driven-Based Infrared False Alarm Source Detection And Classification

Infrared target detection technology is widely used in military and civil fields.In space infrared imaging system,there are many high-radiation objects in the imaging band in addition to the interference caused by the noise of the detector itself,which leads to a high false detection rate and reduces the robustness of detection and monitoring system.In order to meet the requirements of high reliable suppression and intelligent identification of false alarm sources,it is necessary to detect the typical false alarm sources such as celestial bodies and cirrus clouds and distinguish the types of false alarm sources,so as to reduce the negative impact of false alarm sources on threat target detection to a certain extent and improve the accuracy of target detection.The detection and classification methods of false alarm sources in infrared remote sensing images are studied,and the main research contents include the following aspects:（1）The imaging principle of infrared remote sensing image is studied,the similarities and differences of infrared false alarm source in radiation intensity,behavior characteristics and motion characteristics are studied,and the simulation physical model of false alarm source in infrared sequence image is established.（2）The detection method of false alarm source in single frame infrared remote sensing image is studied,and two detection schemes are proposed,which are infrared false alarm source detection scheme based on multi-directional rotating structure operator and infrared false alarm source detection scheme based on local texture feature.In the former,the edge region of the false alarm source is extracted by constructing a morphological structure element,and then the edge region of the false alarm source is connected and filled inside,so as to obtain a complete false alarm source region.The latter plan is to obtain texture feature map by extracting local texture statistical features and local texture spatial features of the image.After image segmentation,the false alarm source region is finally obtained.（3）The detection method of false alarm source in infrared remote sensing image in sequence image is studied,and the method of infrared false alarm source detection based on space-time combination is proposed.The method is planned to get the suspected area of the false alarm source through the difference method by modeling the background,and then use the continuity of the false alarm source track to get the candidate area of the false alarm source.Through the detection algorithm based on the spatial feature extraction of the candidate area,finally,in this section,the trajectory information of the false alarm source can be obtained.（4）The pattern recognition theory is studied in this part,and the classifier of Support Vector Machines is trained by using the characteristics of sample set to realize the discrimination of false alarm sources in infrared images.Discrimination of false alarm source in single frame image by extracting the statistical and texture features of false alarm source region,classifier is used to realize discrimination of false alarm source region type in infrared image.The discrimination of false alarm sources in sequence images is realized by extracting the motion features and gray features of false alarm sources in trajectory information,and then classifier is used to distinguish the types of false alarm sources in infrared images.