| Traditional investigation methods of land utilise have the shortcoming of long renewal period, large workload, low efficiency and high cost. Remote sensing image classification technology can grasp the real data and analysis the land utilization conditions fast and accurately, which has a broad prospect and enormous application value in land use investigation. The traditional pixel-based classification methods, however, depend entirely on the feature of spectral information and ignore the abundant spatial information of high resolution images, which results in the classification results being affected by the metameric substance of same spectrum and salt-pepper noise. In order to solve this problems, the object-oriented classification method is introduced to land survey, as it is good enough to overcome the limitations of traditional methods, and its classification results can be output in the form of vector polygon which can be edit and applicate in COM GIS.This paper use the IKONOS 2010 high-resolution remote sensing image, and choose the northeastern university and its nearby area as the study area. The object-oriented classification method is studied carefully, the main research and contains are as follows:1) Adding texture filtering and edge detection data in the process of multi-scale segmentation, the optimal segmentation parameters can be got. Contrasting the RMAS values of different kinds of ground objects in different scale layers, so that to ascertain the optimal scale for the study area, it finally establish an interconnecting network layer composed of 50,70 and 90.2) In the study, the work area categories are determined by the current national land classification standards, combined with image visual interpretation and field survey. Then establish the classification rule, which have had a carefully analysis of the feature combination, and compared the object-oriented classification results with the traditional pixel-based method. Finally, measure and calculate the length, width and area of feature samples that are seleceted both in the vector map and the field.3) By using object-oriented technology to classify the two phase Landsat 7 images, the change detection rules can be established based on it. Then, the study can extract the changed spot and analysis the statistical area data.Experimental results showed that overall accuracy by object-oriented method is 90.68%, outperforms 18.98% than the traditional maximum likelihood method, apply it to the land use survey, the classification of land use status will be got automatically. However, the measured data in the map are not quite the same with the field as most of the vector results are serrated edges. Thus, the edges should be smoothed before import them into Database. In addition, the change detection method proposed in the paper can detect the land change information quickly and accurately, which have provided an advanced technical means for improving and updating the land use database in time. |
