Research On Rural Mechanically Cultivated Road Extraction Based On Dynamic Weight Constraint

As an indispensable important infrastructure in the process of building high-standard farmland,accelerating the modernization of agriculture and rural areas,and the process of mechanical integration,rural mechanically cultivated road is crucial to improve the efficiency of mechanized production.Therefore,from the perspective of reducing the investment in highstandard farmland construction,it is imperative to effectively extract the information of mechanically cultivated road network.However,the low texture difference between the mechanically cultivated road and the surrounding ground objects,and the narrow road surface of mechanically cultivated road is often blocked by shadows such as crop straw and vegetation,which greatly increases the difficulty of the extraction of mechanically cultivated road,resulting in the weak applicability of the current road extraction method and the decrease of the degree of automation.On this account,this paper takes the high resolution multispectral image as the data source,and uses the geometric linear characteristics and spectral texture characteristics of the road to extract the information of rural mechanically cultivated road.The main work and points of this paper are as follows:(1)Image preprocessing.In order to highlight the image characteristics of mechanically cultivated road.In the preprocessing part,the image fusion,filtering processing,line segment extraction and the determination of the research area of the mechanically cultivated road are carried out to enhance the spatial heterogeneity of the internal and external areas of the roads,and effectively utilize the geometric structure information of the road.At the same time,in order to show the research effect of the method in this paper,according to the third national land use survey norms,this paper effectively selects the images containing the mechanically cultivated road area.(2)Color space selection and measurement function determination.The principle of RGB color space,HSL color space and CMY color space is analyzed.In order to make full use of spectral information of roads,HSL is selected as the color space in this paper from the perspective of similarity indexes selection between roads.By comparing the three similarity measurement functions of Euclidean distance,cosine distance and Mahalanobis distance,euclidean distance is selected as the similarity measurement method of spectral information and applied to the road matching model in this paper.(3)The extraction method of rural mechanically cultivated road under dynamic weight constraint is proposed.The method includes:(1)The Multi-scale Line Segment Orientation Histogram(MLSOH)model is improved from two aspects:(a)The multi-scale line segment constraint is adjusted to a single scale;(b)The length of accumulated multiple segments is adjusted to select the longest segment.According to the angle information of the longest line segment,the local road direction is forecasted and the road prediction direction is updated dynamically in time;(2)Based on the geometric linear characteristics of roads,the dynamic non-uniform weight distribution of different road directions is carried out according to the line length;(3)The multidirectional circular matching template is established,and the dynamic weight distribution and the similarity measure of HSL color space are combined as the matching rules to complete the matching tracking of the rural mechanically cultivated road.In order to verify the effectiveness and universality of the algorithm,this paper selects two GF-2 images with spatial resolution of 0.8m and one Geo Eye-1 image with spatial resolution of0.5m in different regions and scenes as experimental data.The comparative analysis of road extraction results with the other four methods shows that the proposed method has the advantage of higher degree of automation under the premise of ensuring high precision road extraction and vector storage.