Research On Change Detection Based On Unmanned Aerial Vehicles Serial Images

With the rapid development of remote sensing, change detection based on remotely sensed multi-temporal images is expanding in national economy as well as national defence. By analyzing changed information of two registered images of same scene acquired in different times, we can detect the interesting objects which have changed, and use the result to update land environments, track targets and supervise war field. In recent years, the utilization of unmanned aerial vehicles(uavs) and other aircrafts flying in lower air enriches remote sensing platforms, and makes the acquisition of remotely sensed images more conveniently which improves the rapid change detection technology. So aiming at the change detection based on uavs images has important practice meaning and application value.Based on the change detection research inside and outside, this paper summarizes the conception of change detection, analyzes the usual types of change detection methods, then provides a common operation flow for change detection, the thesis focuses on some essential technologies and measures in change detection, mainly includes: feature extraction, image segmentation, relative radiometric calibration and so on. The major works implemented are listed as follows:(1) Because of significant meaning of the man-made objects in change detection, and the flaws of methods based on pixel level mainly concerning image intensity, this paper analyzes the edge features of man-made objects, then uses effective feature extraction operators based on intensity morphology to gain these features, finally exposes a method based on edge features which has been used to discover changed objects and detect them.(2)Aiming at the application of rapid discovering changed objects, the thesis brings forward a change detection method by using the images captured by uavs and spot 5 respectively, the method involves image mosaic, relative radiometric calibration and other pre-processing ways, then by the analysis of misalignment, it presents a change detection method for different source images that can weaken the impact of misalignment to some extent.(3)By analyzing the classical thresholding methods mostly based on image intensity, the paper implements a segmentation means based on fractural dimension which cooperates the intensity information and configuration information to achieve a better segmenting result.Finally, an integrated change detection system for the processing methods recommended above is accomplished in this paper, and the results show the effectiveness of two methods: the change detection method for man-made objects based on homo-source images as well as the change detection method concerning misalignment based on different source images.