Object-oriented Change Detection Method With Critical Technology Research On High-resolution Remote Sensing Images

As the rapid development need of industrialization and urbanization, the exploitation of land resources is growing faster, while a large number of agricultural land are occupied to develop resulted in a significant decrease. These changes in land use are primarily caused by human activities and land remediation. How to control activities on land resources abuse effectively is an urgent issue. Traditionally, land monitoring approach relies on field investigation, or mapping data. These lead to many disadvantages, such as a heavy workload, poor timeliness, lots of labor resources and so on. Thus,understand the work of land use situation periodically is important for sustainable development. Remote sensing technologies are often applied in land use change detection with the advantages of data obtained repeatedly, continuously and large scale ground.This paper summarizes the advantages and disadvantages of existing remote sensing image change detection method and investigates the mechanism from pixel-based and object-oriented change detection. The pixel-based change detection method have shortcomings as process slowly、more salt pepper noise、spectral serious influence. Conversely,The object-oriented detection method can form multi-scales objects and improve detection accuracy greatly by making full use of the characteristics. But there have two difficulty point on object-oriented change detection:①determine the optimal segmentation scale;②choose the best characteristics.Based on the two issues, the main contents are as follow:(1) This paper makes a change detection technical process on high-resolution remote sensing images, which include image pre-processing 、 multi-scale image segmentation 、 image objects feature selection 、 decision-classification and change detection operations.(2) On the problem of the optimal segmentation scale, this paper campares and analyzes the different superior function indicators, put forward an evaluating index including intra-contrast index and intra-entropy index. The paper uses the principle of multi-scale segmentation. The purpose is to get the optimal segmentation scale value by establishing the relationship between segmentation scale and evaluating index..(3)The article studies the optimal characteristics search algorithm of object-oriented method, which is named as Separation Threshold Method(SEath). This paper adopts the search algorithm based on J-M distance criteria to choose the feature. In this basis, the improved algorithm considers the de-correlated feature and the distance in internal class to get the optimal characteristics.(4) With examples of high resolution image, the paper use the optimal characteristics with threshold to classify, build the decision tree classification rules for working out the structured classification rules. The experimental results show that the accuracy of modified optimal feature selection method for classification has significant improvementFinally, this article has experimented on the land use change detection with the classification images. The experimental result has proved that the proposed method is effective and practical.