The Height Of The Cotton Plant Identification Technology For Cotton Top-cutting Machine

Cotton is the domestic essential agricultural cash crop and is of great importance to agricultural economy.According to the data,Sinkiang has been regarded as the widest-planting and the highest harvesting output per acre base of commodity cotton.Sinkiang has also fulfilled the whole-process mechanization in its planting and producing.However during the annual crucial topping period,from early July to the mid every year,the reliance of manual topping decreases the efficiency and increases the labor cost and as a result limiting the scale and the activity of cotton farmers.To increase the efficiency of manufacturing and lower producing cost,it is urgent to carry forward the high-efficiency mechanizing way of topping.There are two restricting reasons for its spread.First,the machine couldn't locate the topping height very quickly or precisely.Second,the rising and falling of the topping knife responds slowly and its command is of low accuracy.Thus,the thesis tries to research and solve this difficulty focusing on the hard identifying problem and is based on the binocular stereo visual technique and image processing relevant theory as well as designing and setting up the visual identifying system after the research into domestic and overseas topping situation and the use of this system.The detailed research outcome and the main process are as following:(1)Collect the statistics and study the growing form peculiarity through a field survey into the growing and planting of cotton.Choose the proper camera lens and the integrating parts,design and build the topping height identifying system based on the binocular stereo mechanical visual theory.Analyze the vision imaging theory and build the model of non-linearity imaging relevant to the calibration of cameras.Then,with the model back stepping,carry out the experiment of Calibration of calibrated board camera to get the intrinsic and extrinsic parameters of the camera.(2)In order to improve the high recognition speed and effect of cotton topping,chooseing the La*b*color space type for the collected cotton images and reduce the dimensionality.Based on the image threshold,the mean processing is adopted,and the image is enhanced without increasing the amount of image data.Using the adaptive threshold method,the color characteristics of cotton and the overall background based on the adaptive threshold segmentation.In order to identify cotton main stem rod,the Sobel operator is used to extract the boundary segmentation,using Hof transform to fit the line of the position of the main stem.For eliminating the interference of leaf,morphological closing operation is used,the accuracy of projection algorithm is optimized and the identified feature of cotton main stem rod.Then based on the binocular stereo visual identification,match and calculate the world coordinate with the identifying point of both the left and the right.Considering the complex conditions in the field,there will be serious occlusion in cotton plant dense planting.Based on the cotton plant identification task,the VGG depth learning model framework is improved,the parameter optimization is completed,and the possibility of realizing the top recognition of cotton plant in the complex field is explored.(3)Design and process the monocase hardware of the cotton top-cutting machine based on the characteristics of its planting and growing.On the base of Dahua Camara SDK,with the help of Visual Studio 2010(for enterprise),design the identifying system of cotton height identifying and rising& falling controlling system.Carry out the simulating experiment of field test in the laboratory with cotton plant,conveyor belt and topping monocase.Finally,analyze and research the collected statistics.Then,summarize the problems in this research,and put forward the potential obstacles in a further research which sets base for the cotton self-adapting topping.