Development Of Hyper-spectral Image Compression System Based On Anomaly Detection

Huge amounts of data transmission and storage problem is a bottleneck, which restrict the application of hyper-spectral image. According to the functional quality of development, we can design the hyper-spectral image real-time compression system. Ensuring the integrity of important information, we can complete data compression and control compression error, which has high theoretical study and social application value. Paper mainly introduces the development of hyper-spectral image real-time compression system, which is based on the anomaly detection of hyper-spectral image, the concrete research contents and results are as follows.(1) Based on the performance characteristics of FPGA, Paper had completed the algorithm design of hyper-spectral image real-time compression system based on anomaly detection with using RX algorithm for abnormal detection, and SVD algorithm for data compression. The algorithm combined the anomaly detection algorithm with rate-distortion controllable data compression algorithm(shared processing), reducing the complexity of the algorithm. The main compute of algorithm is matrix multiplication and eigenvalue decomposition, suitable for parallel running and water processing of FPGA.(2) Using the top-down method and the principle of modular design to complete the design of top structure design of compression system software and the underlying module development. Paper had test and optimized each module through the function simulation, to ensure that each module works stably under ideal conditions. In order to meet the requirements of real-time processing, paper had test and optimized each module through the board test.(3) The paper completed Hyper-spectral image real-time compression system development and compressor performance testing base on PC. The developed system sent raw data to the compressor via Ethernet, received compressed data, and evaluated data transfer rate and compression performance. Test results show that, the Hyper-spectral image compression system can process 400 Mbps Hyper-spectral image data in real-time, without losing any target information, and predicted the mean square error and the signal-to-noise ratio meanwhile. The system can meet the design requirements.