Object-Oriented Change Detection Based on Change Magnitude Fusion in Multitemporal Very High Resolution Images

Very high resolution image shows more detail information of objects on the earth's surface, which makes the traditional pixel-based change detection technique unsuitable for it. The paper proposes a method of object-oriented change detection based on change magnitude fusion in multitemporal very high resolution images. The method can be divided into three steps: 1) Object-oriented image segmentation by multi-scale morphological watershed image segmentation algorithm based on gradient image and marker image. 2) Feature vectors of image objects extraction based on spectral and texture feature. 3) Multitemporal change magnitude maps generation based on change vector analysis, and then the fusion image of change magnitude map can be obtained based on image fusion method, which leads to get change areas by optimum threshold. Experiments show the feasibility and effectiveness of the method proposed by quantitative and qualitative analysis.