| With the continuous development of science and technology, robot vision is widely used in different fields. The automatic lawn-cutting robot can fit a borderline by extracting texture feature of lawn image. In this way it will help us to the working path of automatic lawn-cutting robot in the late study.After the study of image texture segmentation, the gray level co-occurrence matrix is used to extract the texture feature value of the lawn and the texture feature is mapped to a binary image which shows off the texture feature. According to characteristics of human vision the breakpoints in each transverse which had the same width can be determined. Finally these breakpoints will be connected with a borderline. The main work are as follows:(1) Research of texture feature extractionThis paper discusses the significance and the related fields of image texture feature. The extraction technology of texture feature is analyzed in detail from four aspects: model, statistics,spectrum and structure. The experimental results of several algorithms are compared.(2) The extraction of the binary image based on texture featureThe algorithm of gray level co-occurrence matrix is applied to the texture segmentation of the lawn image. By comparing the four different matrices of eigenvalue, "Correlation", "Contrast", "Energy" and "Homogeneity", and the four different traversal windows, "3×3", "5×5", "7×7", "9×9", as well as the four different gray levels, "8", "16", "32", "64" this paper selects the best segmentation parameters which can extract the best effect of texture segmentation from the binary image. The experimental results show that on the basis of selecting the appropriate parameters using gray level co-occurrence matrix can extract the texture feature of lawn image quite well.(3) The extraction of the borderline based on texture featureBy use the logic judgment of human eye when it distinguishes the borderline of a binary image, this paper proposes a algorithm of borderline fitting based on the texture segmentation image of lawn. Firstly the binary image should be divided into a plurality of aequilate transverses.The size of the contour traversal can be determined by calculating the occupancy of black pixels in a single transverse. Then according to dispersion ratio of the black pixels the the length of contour traversal should be increased appropriately. About the extra length this paperproposes the increase by fixed value and the increase by changed value and has compared these two methods. Afterwards all individual transverse must be traversed by using the determined contour traversal in order to make sure the position of breakpoints. Consequently, these breakpoints are connected with straight lines. The experimental results show that the algorithm can fit segmentation points accurately and distinguish between the image of the worked area and the non working area.(4)The path planning based on borderlineAccording to the borderline form the fourth chapter we research for the path planning of Robot Mower. Firstly the position of segmentation points in the image are projected into the real areas. By analyzing the difference of angle and distance between each segmentation point we try to calculate the rotating angle and mowing distance. Eventually the MCU controls the stepping motor to complete the final path planning.Key words: Lawn segmentation; CLCM; Eigenvalue matrices; Contour traversal; Path planning... |
PDF Full Text Request