Building Texture Acquisition And Processing Based On Unmanned Airship Low Altitude Aerial Survey System

Since the America former Vice President AI Gore put forward the concept "Digital Earth" in1998, many countries in the world generally make the city as one of the main entry points. The theory and research framework of "Digital Earth" have been the hotspot for Geographic Information Science.With the rapid development of Grand Integration and Cloud Computing, the demand of "Smart City", as a new generation of information support, for the high resolution remote sensing data is growing with each passing day. The primary problem of national construction and informatization is how to quickly and accurately obtain the basic geographic data.In recent years, low altitude remote sensing technology is developed quickly, which overcomes the insufficiencies in timeliness and accuracy (The image accuracy includes centimeter level precision and3D building accuracy) for satellite remote sensing and high altitude aerial. The existing low altitude photogrammetry is most for light small fixed wing UAV with a single digital camera. It has the advantage of light energy of photometry, but also the weakness of geometry of narrow image, which not only makes the low work efficiency, but also leads to decrease the precision of aerial survey.In order to solve the problem of low accuracy, the basic way is to use a wide angle camera. But the domestic and foreign mature wide format combined camera lacks the objective foundation for using in the unmanned air vehicle platform. On the other hand, because of the restrictions in the flight height and photographic pose restriction (vertical or small angle photography),it only gets the height information and top texture information of part of main building, which is difficult to meet the requirements of data acquisition for3D city building.Combined with the newly development of multi-sensor integration technology, this paper researches the wide format and wide viewing angle aerial camera used on the low altitude aerial, which can allow for quick and safe3D city building information acquisition to establish a realistic3D city models with texture.The main research contents are as follows:(1) The selection of low altitude aerial flight system with large scale and design of the sensor system are done. Firstly, the influence factors of low altitude aerial with large scale to obtain multi-view images with high resolution and large scale measurement accuracy is analyzed, which is as the basis for the selection of low altitude aerial flight system with large scale and design of the sensor system. We can see that, in order to obtain high resolution and high definition images, the aircraft must fly slowly and lowly. So the low altitude aerial system with unmanned airship is selected as the remote sensing platform to obtain3D city data source. Its advantages and system design are discussed. Then, the light small wide angle combined camera system for unmanned airship is designed, including texture camera model design and stable platform design.(2) The accuracy of low altitude aerial with texture camera is analyzed and the method of improving the accuracy is researched. Firstly, the accuracy problem associated with wide angle imaging is analyzed, including the relative orientation accuracy and mapping accuracy. We can see that, the low altitude aerial accuracy has a great relationship with the image field angle of the camera. Therefore, it is as far as possible to use wide angle camera for low altitude aerial. Then, how to improve the aerial accuracy by the designed texture camera for UAV is discussed, including effective image judgment, self-calibration self-stabilization technique and the operation method of using large overlap images. Finally, the test is done to prove that low altitude aerial using wide angle camera with large overlap can effectively improve the aerial accuracy.(3) The method of texture extraction by texture camera is researched, including automatic fast method and semi automatic method for building contour and images registration. Firstly, according to the building boundary by linear expression, how to accurately obtain the linear characteristics of the building edge on the images is the key to improve the texture extraction. The automatic fast method of texture extraction, that Canny algorithm is used for edge detection and the weighted confidence is used to obtain the best image edge line by the angle and distance constraints, is put forward. Then, considering the complexity of registration between the building space contours and image linear feature in multi images, the possibility of semi automatic method to improve the registration relationship by the linear feature matching is discussed. The matching strategy of image linear features with multi view images based on geometric constrains (Epipolar constraint, three optic tensor constraint, similarity constrain) is proposed. Finally, the selection algorithm for multi texture images and geometric correction processing is researched.(4) The optimization method for building facade texture is researched, including texture images deblurring and adjacent image colors consistency processing. Firstly, the reasons for the fuzzy phenomenon of texture images is analyzed, including the effect of noise, image shift and interpolation processing for geometric correction etc. Based on the advantages and disadvantages in comparison to existing image restoration algorithms, the MTF image restoration algorithm based on high frequency emphasis is proposed. The MTF cure is estimated by the image edge data, which can avoid repeated PSF estimation process. The effectiveness of the proposed method is proved by the image quality evaluation standard. Then, based on the phenomenon that a single building exists mottled colors in various surfaces and adjacent images colors transition is not nature in the construction of the building texture, the main color matching method is proposed based on the texture camera images. This method can correct the dark or bright images and make the texture image tone basically the same to achieve the3D city visual consistency and aesthetic effect.