Study Of High Quality Data Acquisition And Processing Technology For Airborne Multi-spectral Camera

In recent years, with the development of spectral technology and the increasing in the application of spectrometers, the spectral resolution and spatial resolution of spectrometers have been greatly improved. Especially in remote sensing field, higher quality of spectral data is needed to meet more strict demands, and with the sharply increasing of spectral signal quantity, signal the number of spectral information increases sharply, high-speed acquisition of high quality spectral data is necessary to meet the demand of applications. Thus, it will add to huge pressure on the processing of spectral data, especially in the field of spectral data real-time monitoring which needs more rapid offline hardware processing platforms. In this paper, most of all the important aspects of airborne large filter-array-type multi-spectral camera imaging chain were studied to make multi-spectral camera with miniaturization, low power consumption, high performance and real-time data processing become to true. After successful its flight test, multi-spectral camera got high quality data, and entered the phase of mass production. The main research harvests are as follows:1. An imaging model and the critical parts of the imaging chain of the filter-array-type multi-spectral camera were constructed. The engineering design of multi-spectral camera, the main part of the imaging chain, was completed, which highlights it’s miniaturization.2. For the first time, the low power bi-stable mechanical electronic shutter served as lens shutter was used in aviation large array multi-spectral camera. Along with introduce to various parameters of the shutter, the impact of the shutter on spectral data quality of the large filter-array-type multi-spectral camera was analyzed. Based on the analysis above, the paper proposed an optimizing scheme in the mechanical structure and electronic drive etc., and proved the validity of the improvement through indoor experiment and outdoor push broom test. With the improved design strategy, the maximum rate of change between adjacent exposures was reduced from 15.05% to 0.96%, the combined color images’ color standard deviation was reduced from 2.0455 to 0.3738%, therefore, the authenticity and consistency of the target spectrum wasenhanced, and provided high quality image data for subsequent processing of multi-spectral data.3. The paper do the work of laboratory spectral calibration, relative radiometric calibration, absolute radiometric calibration and the uncertainty of all the calibrations were analyzed for the filter-array-type multi-spectral camera to eliminate the non-uniformity imaging caused by optical system and CCD detector, improve the quality of multi-spectral image data, achieve spectrum quantitative detection. In the experiment, we found the mechanical shutter’s instability cannot be eliminated and has a great effect on the calibration results. Therefore, according to the imaging characteristics of the spectral camera, this paper put forward a method of laboratory calibration methods without mechanical shutter, which bring about a great improvement in the precision and stability of calibration coefficient.4. To realize the image motion compensation of the filter-array-type multi-spectral camera, we built airborne two axis stabilized platform and utilized the improved PID and lead-lag compensation control method. And during the push broom test, we got a high quality multi-spectral image, which verify the effective of the platform and method proposed above.5. Combined with the filter-array-type multi-spectral camera principle and data processing character, this paper present a method of registration before mosaicking to improve the result image registration accuracy. This paper also present a multi-spectral camera data generation method by automatic extraction of pairs of homonymous points and lines, the process of which mainly includes: automatic homonymous feature extraction, geometry transform parameter using adjustment calculation, registration and mosaicking. A matching method for multi-spectral images called ‘improved Hough transformation with disparity constraint’ was presented. In the matching strategy, SIFT operator is used to the top layers of the pyramid imaging to provide the initial disparity constraint condition. As to the other layers, improved Hough transform method is applied to match the left and right pyramid images. In this process, the speed and accuracy of image matching is improved by using upper matching results as the disparity constraint conditions of the current layer image matching.6. The spectral image target detection and online fast parallel processing technology was studied. By contrasting spectral unmixing matching method with abnormal target detection, the latter one was verified more suitable for online processing and algorithm transplantation. Then we focus on the RX detection algorithm, and an improved algorithm called kernel RX which is suitable for engineering practical is presented. By studying the parallel characteristics of RX algorithm, a flexible FPGA and multi-DSP parallel processing hardware architecture and test platform was built to implement and transplant the parallel algorithm. The experimental result shows that the platform can realize the spectral data fast online processing and result real-time transmission, which satisfy the requirements of the UAV airborne spectral data fast application.