Research On Change Detection From UAV High Resolution Images

With the development of social economy and science and technology, all kinds of activities of human bring about changes to the earth---the form of landscape, in particular, with the development of urbanization and new rural construction, great change has taken place. In addition, the crustal movement, volcano, earthquakes and other natural disasters, the natural environment and surface morphology vary from minute to minute. In order to monitor the surface changes timely and effectively, it is of great importance to analysis of the characteristics of surface transformation, update the relevant geographic information systems so that we can provide scientific and effective reference data for resource management and planning and environmental protection and other relevant departments of the decision-making. Change detection based on UAV images can improve the efficiency, so the systematic research on UAV image change detection and its related technology are of vital important in order to updating the geographic information system data timely, finding and forbidening the illegal land, illegal exploitation, abuse of farmland, the destruction of the ecological environment in a much more efficient way. Rapid response to natural disasters are of great significance.According to the characteristic of the UAV images, we put forward relevant technology of UAV image change detection. Making full use of the advantages of UAV images, trying to avoid UAV image defects, systematic analysis of the related theory of UAV image change detection technology is be fully investigated.The paper focuses on solving the problems as follows:1?Borrowing some idea from traditional change detection methods, we introduces the change detection process and the challenges encountered, meanwhile, the factors that affect the accuracy of change detection is presented. By analyzing the characteristics of UAV images, preprocessing of UAV images was studied.2?The preprocessing of UAV images is investigated, which include radiometric correction, shadow detection, geometric registration steps, and then detailed description of each phrase is given. A shadow detection method based on saturation of colour is proposed, experiment result shows the validation of this method. Further more, in order to improve the quality and quantity of matching, a new SIFT feature matching method is proposed based on graph theory.3?Taking advantage of the high resolution of UAV image change detection, wavelet transform theory is employed to extract the multi-scale features, and MRF theory is used to proposed a change detection method based on multi-scale fusion. The effectiveness of the method is verified based on both simulated and real data sets.4% Due to the fact that change areas usually result in the changes in other features such as regional area, perimeter, shape, texture and etc. A change detection method based on affine invariant feature and image segmentation is proposed, which take the extracted but not matched SIFT points as candidate changed area.And then, the result is refined by considering other information such as intensity of region.5?The UAV image usually suffer from serious projection and registration errors, on the basis of study of local invariant feature and object change detection mechanism, a novel change detection method based on local invariant features and image segmentation is proposed. Experiment shows that our method outperforms other methods and can weaken the influence of image registration error on change detection to a certain extent.