A Research On Dual-band IR Object Simulation And Its ATR Method

The IR ATRs based on infrared objects simulation uses techniques of computer graphics, digital image processing, heat transfer to simulate IR images, and uses pattern theory to complete object recognition. The simulation and recognition of IR objects is one of high-techniques in military, so such a detailed research is theoretically meaningful and of extensive prospect of real applications.This paper discusses techniques for scene simulation based on 3D model and issues on system development. We also give a new idea for ATR algorithm design, i.e., a "top-down" approach for ATR algorithm.The development of IR imagery techniques gives IR techniques, which have been widely used in military, a rare chance; at the same time, the come-forth of SLSI and super PC, makes it convenient for the test and adjustment of IR intelligence system by using the simulated IR imagery. In this paper, we use the Wavefront OBJ file data, which is the output of commercial software Maya, realize the dual band IR imagery simulation with high detail based on 3D model. The data is useful in ATR algorithm. On modeling, this paper probes into the rendering effect and efficiency in the following ways in order to get better simulated images: (1) In calculating of the temperature field, I used the "point source" heat model and "slice source" heat model separately, and also analyzed and compared the two models calculate efficiency; (2) In the use of "slice source" heat model, this paper compared and analyzed the calculation methods of form factor. The paper also gives a detail introduction on how to use Visual C++ to read 3D models and how to combine Open GL to rendering the models. In the end, after presented the object model in certain condition, the paper also gives the simulated images using the way I discussed above. Then, on the basis of the images, in addition to "point source" methods and "slice source" methods, the simulated results of dual band IR images are explained and analyzed.In the previous decade, domestic and overseas researchers have made a thorough research on the ATRs of IR objects. As a complicated system, an ATRs may meet many kinds of problems. The main difficulties involved in ATR are as follows: (1) Geometric variability: Targets may take on different appearances depending on the orientations and positions of the object; (2) Image variability: The data obtained may vary with environmental conditions and will be subject to the noise characteristics of the sensor, in addition, the textures comprising a scene will be different;(3) Scene variability: The objects number is unknowable and they can also enter or leave the scene. To be a robust ATR system, besides expressing the flexible and real the natural scenes, invariance to the above three points is required. The traditional recognition makes usage of machine vision techniques by preprocessing (edge detection, segmenting), and then conducts a higher level inference (character extraction, classification, recognition); In recent years, some foreigners combines the machine vision and graphics to give a "top-down" algorithm, this method uses the detected data directly, join the knowledge of objects to make inference.