Research On Automatic Extraction Of Terraced Ridges In Loess Hilly Region Based On GF-2

With the rapid development of remote sensing technology and image extraction technology,the method of obtaining terrace information based on images has become more and more popular,and has achieved certain results.But people ignore the status of the number and area of ridges in land use,especially in cultivated land.The larger the ridge coefficient,the smaller the actual arable area of the arable land.The study of the ridge coefficient has an important role in protecting cultivated land,saving intensive land,and measuring soil erosion in terraced areas.This study is based on the GF-2 remote sensing image with 1m resolution,using multi-scale segmentation of remote sensing image and object-based information extraction method to explore the accurate and rapid met hod of extracting the ridge information of loess hilly and gully region,and draw the following conclusion:(1)In this study,through analysis and determination,a multi-scale segmentation test was carried out in a trial-and-error method to determine the optimal segmentation scale parameter settings for the related features of terraces and to construct different image object layers.It is determined to extract the optimal texture features with four feature indexes of angular second moment,entropy,contrast and correlation.In order to improve the accuracy of image classification,after analyzing the feature information of the subject,spectrum,shape,texture and other objects formed by image segmentation.Using the optimized feature space as a reference,determining the features or combination of features that are separate in each place,constructing the feature space and classification rules for each land use type,and achieving different levels of each place classification.(2)On the basis of extracting the terrace area,using the method of hierarchical segmentation and extraction,taking advantage of the characteristics of the small density of the ridge,constructing the ridge extraction rule with Density as the main feature information,which realizes the rapid and accurate automatic extraction of the ridge information.And calculate the ridge coefficient as 15.34%,and finally evaluate the accuracy of the classification results,the overall accuracy of the classification is 82.85%,the Kappa coefficien t is 0.79,the classification quality is good.The accuracy of the terrace extraction results reached 82.55%,the Kappa coefficient reached 0.75,the field extraction accuracy was 68.83%,and the Kappa coefficient was 0.61,which achieved a good classification effect.(3)A simple random sampling method was used to select the terrace samples,field measurements and high-resolution UAV image visual interpretation were used to calculate the ridge coefficients,respectively,and their validity was verified as the extraction of the ridge coefficients.The calculated sample sizes of the two methods are 10 and 11,respectively,which is lower than the pre-sampled sample size of 20,which can be used to verify the accuracy of extracting the ridge coefficient.The comparison results show that the automatically extracted ridge coefficient is 15.27% higher than the field measurement ridge coefficient and 14.04% of the UAV image visual interpretation of the ridge coefficient is 0.07% and 1.3%.The analysis believes that it is the effect of image resolution and shadow.