Scale Estimation Of Object Oriented Image Analysis Based On Spectral Spatial Statistics

Multi-scale segmentation is the key step of analysis of remotely sensed imagery.Scale parameter choosing in segmentation process is directly related to the quality and accuracy of object-oriented analysis.Only on the basis of experience for segmentation parameter choice that has less quantitative analysis ways,currently.These methods lack of quantitative estimation before segmenting,with large workloads and low efficiency.From the perspective of scientific research,its scientificity and universality is poor.The objective of this paper is using the method of quantitative method to determine the scale parameter.In order to realize the automatic extracting target on object-oriented.This paper summarizes the concept of scale parameter in object-oriented analysis.This paper analyzed the commonly used segmentation scale parameters into spatial bandwidth and attribute bandwidth.Methods this thesis used a spatial and spectral statistics-based scale parameter selection method for object-based information extraction from high spatial resolution remote sensing images.The relationship between Fractal Net Evolution Approach(FNEA),Mean Shift in multi-scale segmentation and spatial statistical characteristics was analyzed.Scale estimation based on spatial and spectral statistical characteristics is applied to FNEA in multi-scale segmentation.Meanwhile,the scale estimation approach proposed in this paper is verified by high spatial resolution image——IKONOS and SPOT5 data.Construction area and farmland area were selected for spatial and spectral statistical characteristics to further estimate the optimal scale parameter in segmentation.To verify the reasonability of the predicted optimal scale parameter,a series of supervised classification was performed.The classification results and the accuracy assessment results shows that the estimated scale by spatial statistical characteristics is basically closed to the optimal one in FNEA or Mean Shift based multi-scale segmentation.The proposed scale estimation approach can ensure the accuracy of the following object-oriented image classification.The method can be used to estimate the appropriate scale parameters before segmentation.In addition,it is an essentially data-driven method that requires almost no prior knowledge;thus,it can enhance the efficiency and automatic degree of Object Based Image Analysis.