An Image Reduce Scheme And Key Algorithm For Air-ground Data Link System

RS(Remote Sensing) is an emerging discipline that has bright future of wide application. So far it has been the only way that can provide globally dynamic data observation. The main features of remote sensing are special continuity and time sequentiality. Remote sensing image data is widely used and military usage is one of the most important.There is profound development of aerial remote sensing in China and data of different scalability and temporal can be acquired. With the enrichment of aerial platform and the increasing demand on real time data transmission, there is an urgent need to construct a serial of air-ground data links in order for different aerial platform to perform different task. But for different aerial platforms there is much difference in data volume and transmission condition, airborne equipment volume and weight limitations. All these must be considered when construct air-ground data links.UAV(Unmanned Aerial Vehicle) is widely used in military scouting and monitoring. These applications require high-resolution 2-D or 3-D images that can describe geometrical shape of the object. For UAV, real time transmission of the findings are essential especially when perform important military scouting tasks. So air-ground image transmission technique became increasingly a hot spot for more and more researchers.For modern air-ground data link communication, with the wide use of such sensors as airborne radar and CCD airborne camera, the data volume became larger and larger but the transmission rate of the data link is much lower than the maximum rate of the sensors. The difference in transmission rate makes it impossible for the original data form sensors to be transmitted to the ground, and an airborne processor is needed to lower the transmission rate to an acceptable level.There are two methods that can be used to reduce transmission rate: data compression and data reduction. Data compression can transform data to a moreefficient format. Almost all information is reserved after transformation and the original data can be reconstructed on the ground upon requirements.For still image acquired by CCD airborne camera lossless compression is adopted in order to preserve maximum information of the object. But when the data information is too larger to bear, lossy compression with little distortion is also considered. Because there is a limit in bits representing the acceptable image, so once the BPP(Bits Per Pixel) is lowered as much as possible, the number of pixels to be transmitted must be reduced accordingly. To achieve this data reduction instead of data compression is needed. Data reduction is a process to discard some data to lower transmission rate and of course, some information is lost in this process. But if it is done in an effective way, the information discarded is useless information in performing the tasks, then it has slightly or even no effect on performance.In the process of aerial shooting using CCD camera by high altitude UAV and manned scout that equipped with image transmission sub-systems, there is strong relations between the two neighboring images because large overlapping and strong correlation between neighboring images acquired by multi-angle imaging. When performing real tasks, there must be about 30 percent or even 55 to 65 percent of course overlapping between neighboring images. Because different tasks entail different requirements, so there is no sense in down transmitting these overlapping scenes when the tasks demand immediate acquirement of battle field information.To satisfy the requirement of real time image transmission and take into consideration the bad adaptive ability of air-ground data links to different aerial platforms that performs different tasks, according to the actuality and development rend of data links, based on the characteristic of multi-angle remote sensing images and information extraction methods, have a better understanding of performance of operation tasks, design different operation flow for different tasks in order to make most of the limited transmission rate, thus increase image transmission efficiency of data link. On the basis of requirements of practicalscouting tasks, an airborne fast and automatic image data reduction scheme is presented in this paper. The algorithm fist perform clipping on the overlapping scenes between neighboring images, then it only needs to transmit the matching parameters between neighboring images and the clipped images. The proposed scheme reduces the data volume to be transmitted through the downward links and thus raises the efficiency of air-ground data links and RS image processing systems.The manned scouts usually fly on an altitude more than 10,000 meters when perform military aerial shooting tasks, high altitude UAVs usually fly in the stratosphere which is about 20,000 meters above the sea level. CCD sensors use central projection imaging technique, and the automatic steer on the airplanes and the automatic servo mechanism of the aerial camera can also ensure that CCD sensors automatically and stably perform vertical shooting. So 2-D affine transform is enough for average conditions. The paper uses the phase correlation operation of $MT(Fourier-Mellin Transform) to adaptively obtain matching parameters of visible light images, thus automatically realized image shift, rotation and scale. Adaptive parameter acquirement is the critical step in the implementation of automatic clipping.Among various image matching techniques, the area based ones such as correlation-like and mutual information usually have large computation burden and low speed. When there is a small angle rotation or a small scale in the matching performance is lowered accordingly. Nowadays the most studied method is that based on features sets, for example, relaxation method, invariant descriptor method. But these methods have weak robustness and when error matching rate is high, they need interaction between man and machine and can't satisfy the requirement of totally automation. Pyramid algorithm based on wavelet modulus maximum edge detection and normalized cross-correlation matching is a hot researching spot but it is very complicated and its performance also has limitations.Analysis and comparison show that for image matching, FMT is a verypowerful mathematical tool because the transformed spectrum is invariant for rotation, shifting and zooming. And because Fourier transform has lots of mature lifting algorithm and it is easy to implement on hardware, it has some unique advantages in implementation.Phase-relation matching algorithm based on FMT combines Fourier transform and log-polar coordinates format Mellin transform and it is a rotation, scale and shift invariant shape matching technique. This technique uses the invariant characteristic of the respective operators. Fourier transform can acquire shifting parameters, then in log-polar coordinates system change the rotation and scale operator into shift format, and through phase relation the rotation and scale parameters can be achieved. At the same time, FMT also improves on the speed of phase-conelation method. Its excellent performance in complexity and robustness make it the ideal choice for the implementation of fast and automatic image clipping algorithm. In light of this it is adopted in this paper.After acquiring the shift, rotation and scale parameters of neighboring images, highlight object detection is performed in overlapping parts to ensure that the discarded information is useless in get the tasks done. Image difference method is used in this paper. If no object is detected, clipping points of the reference image are selected according to clipping precision requirements. The clipping line is introduced form clipping points to the direction of rotation. The position of the pixels contained in the area to be clipped is identified through clipping lines. Give these pixels value that represents "white" but remember that the size of the image is not changed. Up to now the clipping operation on the reference image is done.This paper focuses on aero-photograph that is used by high-altitude-UAV and manned scout to fast acquire battle field information. An automatic image data reduction scheme is presented to adapt to the developments of modern air-ground data link, especially long-distance data links. It provides a new way to design task operation flow. But the scheme touches on many issues which can't be covered in only one paper, so this paper only studies the most important part, that is,...