Research On Lifting Control System Of Mower Cutter Based On Binocular Stereo Vision

With the development of society and the improvement of people's living standard,the greening level of the city has become the focus of attention,and the consequent problems of lawn management and maintenance need to be solved urgently.The diversity of lawn species leads to different maintenance and management schemes,and the main lawn stubble height standards are quite different When the ZTR lawn mower maintained in large grassland,it's cutter can't raised and lowered automatically and the initial height of lawn plants can not be accurately measured.So the precision control of stubble height can not be realized in mowing operation.In view of the above problems,lawn plants are classified and their height are measured based on computer vision in this paper.On this basis,the lifting height of the lawn mower cutter is calculated automatically,which lays a foundation for the automatic height adjustment of the cutter.The main contents are as follows:1.A measurement system for binocular stereo vision of lawn machine based on Halcon is built.According to the principle of binocular camera imaging,the binocular camera is calibrated by self-made calibration plate.The results show that the calibration results are stable when the number of images is more than 25.2.Lawn plants are identified and classified based on LS-SVM and multi-feature fusion.Firstly,considering the sensitivity of HSV color space to local color changes,two pixel-level color features are extracted by Sobel operator.Secondly,three texture features of lawn plants are extracted by Gabor filter,and seven shape features are extracted by region correlation operator.Finally,the training samples of LS-SVM classifier are selected by dynamic threshold method,and the penalty coefficient C and the kernel parameter ? are determined by Bayesian optimization method.The experiment shows that the accuracy of lawn classification is 92.88%,and the classification algorithm can meet the requirements of different weather.3.Measuring the height of lawn plants after recognition and classification based on the established binocular vision system Firstly,the traditional Foerstner feature point detection algorithm is improved,which not only strengthens the edge information of lawn plants,but also reduces the number of false corners.Then,in order to reduces the amount of matching calculation and improves the matching accuracy,the extracted feature points are taken as matching objects,and a stereo matching method combining sequential constraints with dynamic programming is adopted.Finally,the virtual ground is constructed based on the coordinates of the feature points.The average height of the plant was obtained by point product operation between the spatial coordinates of the characteristic points at the top of the leaf and the normal vector at the bottom plane.The experiment shows that the error rate between plant height measured by the above method and its actual height is about 1%.4.The feature database of lawn species was established based on SQL.The color,shape and texture of the turf are managed in a unified way,and the relevant parameters such as mowing frequency,period and stubble height were added to the management of turf maintenance.Finally,a perfect lawn feature management system is established to lay a foundation for the development of intelligent lawn mower.