Research On Terraced Field Extraction Method For UAV Image And Slope Data

Terraced fields are terraced sloping cultivated land formed by human transformation of natural terrain.It is an effective measure for soil and water conservation.Because the distribution information of terrace area and block number is not easy to get quickly,it is difficult to carry out quantitative research on the role of terraced fields.With the continuous development of UAV technology,it is possible to acquire high-precision terraced terrain information.Based on the unmanned aerial vehicle(UAV)Orthophoto Image and the slope data,the BP neural network method,the SVM machine learning method and the improved Canny edge detection method are used to separate the terrace area,and the number and area of the fields,the number and area of the fields in the terraced field are extracted by the region growth algorithm.The statistics of information.The main contents and results are as follows:(1)Segmentation method of terraced fields based on neural network.In this paper,a terraced field segmentation method based on BP neural network is used to obtain the information data of terraced fields by learning the model of the feature rules between different ground objects and then dividing the terraces.Experiments show that the algorithm is better than the 1 type of terraced fields for the 2 types of terraced fields.The average integrity and error rate of the method are 73.78% and 28.56% respectively.(2)Segmentation method for terraced fields based on support vector machines.The trusted training sample points are obtained by artificial selection,and the unified unified kernel based on the Superpixels-based Classfication via Mutilple Kernels(SC-MK)algorithm is used as the classification criterion of SVM.Finally,the terraced field results of the experimental sample area after the SVM segmentation are given,and the leveling is calculated.The average integrity and error rate were 83.12% and 43.16% respectively,and the method was more suitable for segmentation and segmentation of 1 types of terraced fields in terraced fields.(3)The improved method of Canny edge detection in the terraced area is designed and realized.In view of the problems of incomplete field in the field of terraced field and the existing holes in the field,from the terrain features of the terrace,the DEM data and the unmanned aerial vehicle image data are combined to realize the segmentation of the terraced field.The average integrity and error rate of this method are 87.72% and 11.4% respectively,although the method is the most complete to the terraced area compared with the SVM and BP neural network method,but the edge of the terrace is relatively rough.(4)Field extraction method based on regional growth,based on the segmentation of terraced fields,in view of the inability to further obtain the information of the number and size of the terraced fields,a regional growth algorithm based on graphics is proposed to fill the area of the terraced field,and the staircase field is included in the image.The number and area information of the block are compared with the results of the visual interpretation of the terraced field.It is verified by the experiment that the accuracy of the method is 83.57%.Because the method can finally make a direct statistics of the number and size of the fields in the terraced area,it will be more beneficial to the digital mapping and mapping,and it is a successful exploration of the terraced field extraction algorithm.