Infrared Target Recognition Technology On Coastal Scene

This thesis proposes a method to address the problem of infrared objectrecognition on coastal scene. Coastal infrared images are simulated utilizing theinfrared radiation characteristics of coastal scene and an algorithm for infraredobject recognition is proposed.This thesis introduces a real-time infrared scene simulation approach basedon texture temperature correction. On the basis of traditional infrared modelsimulation approach, this new approach is used to obtain infrared scene byassigning fundamental temperature filed to a specific kind of material throughenvironmental modeling, revising the fundamental temperature field by usingtexture luminance under visible light of this environmental model, calculatinginfrared radiation with principles related to radiation.This thesis utilizes OSG(Open Scene Graphic)3D rendering engine, optimizes infrared radiationtransmission process modeling by GPU programming technology, obtain sreal-time generation of infrared simulation scenes under complex background,and finally verifies its simulation effect through testing.This thesis proposes to utilize the HOG feature and linear discriminativeanalysis to recognize coastal infrared objects. The HOG feature is robust underdifferent light conditions. Multiple templates are trained for different scales andviewpoints, which improves recognition accuracy effectively. We build a newinfrared ships dataset which were labeled and used in our experiments.Performance under different settings are demonstrated.Finally, we build a maritime optical combat simulation system byintegrating the infrared simulation technique and the infrared object recognitionalgorithm. The simulation system is speeded up by improving the imagesgeneration module and display optimization.