Research on Joint Target Detection for Dual-Sensor Image and System Implementation
|School||Harbin Institute of Technology|
|Course||Information and Communication Engineering|
|Keywords||multi-sensor image fusion target detection evidence theory|
With the development of modern science and technology, the technique for single sensor information processing has been widely applied in the battlefield surveillance, industrial process controls, remote sensing, image processing, and many other areas. This technique has played an increasingly important role in both civilian and military fields. However, with the help of single sensor, people can only obtain a certain aspect of the information of a target or a scene; cannot make an accurate and comprehensive presentation for the target or scene. Under such a background, the multi-sensor information joint processing technique came into being. The advantages of joint processing of multi-sensor information are as follows: with the help of this technique, comprehensive information of a target or a scene is obtained, and through analysis and composition of that information, consistent presentation of the target is obtained so that the accuracy and reliability of conclusion are improved. The use of images from different sensors to conduct joint target detection is not only a popular application of joint processing technique of multi-sensor information, but also one of the focuses in the area of image processing at home and abroad in recent years.In the joint target detection of multi-sensor image, the results of target detection for single sensor image will have a direct impact on the integration of the final joint detection results. So the research in the target detection algorithm of single sensor image is essential. The focus of joint target detection of multi-sensor image lies in how to appropriately composite the information from different sensors. Researchs on SAR image and hyper-spectral image are made as follows:Firstly this thesis focuses on the double parameter CFAR detection algorithm of SAR image, as well as the RX detection algorithm of hyper-spectral image. The experiments show that these two algorithms perform well in single image sensor target detection, but their shortcomings are also obvious.Secondly, on the basis of deeply research on D-S evidence theory, a joint target detection algorithm for dual sensor is proposed. The experiments show that under the condition of 10-3 magnitude as false alarm probability, the detection probability of joint target detection algorithm reaches 84.51%, increased by 11.27% and 19.72%, comparing to target detection only using high-spectral image and SAR image.Finally, on the basis of algorithm research, a joint target detection system for dual-sensor image is designed and implemented. This system is mainly composed of the data source simulation module and joint target detection module. The data source simulation module take FPGA as the core, is mainly responsible for image data transmission and storage under different state; the joint target detection module take DSP as processing core, mainly implements the hardware completion of joint target detection algorithm for dual-sensor. Results of the system show that the speed of USB interface reaches 19MB/s, and it takes 17.3s for the hardware system to work out the result for the joint target detection. The detection probability of joint target detection algorithm increases by 12% at least, comparing to target detection using image of only one sensor.