Cotton Plant Recognition System Based On The Machine Vision Technology

Xinjiang is China’s largest high-quality cotton production basement. The total and theyield per production of cotton, has ranked first for so many years in the country. With theincreased efforts to develop the west of the country as well as the comprehensivedevelopment of Xinjiang, the strategic position of cotton in Xinjiang is more prominent.Cotton topping control is a key link in the process of the harvest of cotton production, whichlast from late June to early July in each years. But due to the problem of hard to identificationthe cotton features at real-time and accurately for some of the domestic cotton toppingmachine, the driver can’t timely adjust the height of the cotton topping according to thegrowth of cotton, which cause the phenomenon such as un-topping, hit the bell, hit the peach,damaged cotton leaves and so on. The cotton topping in Xinjiang is mainly by manual. Theresearch to solve the problems of automatic identification of cotton when cotton topping, is akey challenge to realize the cotton topping accurately, but so can achieve the main steps ofcotton mechanization of the whole. This paper post a cotton plant identification system, whichbased on machine vision technology, the mainly technology used and the contents are asfollows:1. Study the current situation of the domestic and foreign cotton topping machines,cotton plant identification and machine vision technology by system analysis method. Postresearch contents, research goals and technical route after the analysis of the existingproblems.2. Design the static machine vision systems and the real-time machine vision systemswhich use the theory of mechanical system design and machine vision theory, combined withthe actual situation of cotton fields, cotton plant. Complete equipment selection, installationand debugging for the experimental platform.3. Analysis the basic principle of camera calibration, complete camera calibration by theuse of the checkerboard image and Jean-Yves Bouguet camera calibration toolbox, finish thecalculation of the left and right camera projection matrix.4. Collect the cotton and cotton plant images according to the research target and theseason, complete the preprocessing of the cotton and cotton plant images by using imageprocessing technology.5. Study the segmentation results obtained by implementing in different color spacemodels for cotton top. Extract the cotton top image and its features.Use artificial neuralnetworks to achieve top recognition.6. Use the method of feature points matching to realize the stereo image matching of leftand right cotton plants. Use the calibration parameters to achieve the prediction of the cottonplant depth information. Analyze the causes of error. Establish the prediction model for cotton depth prediction which based on the image.7. Use the method of regression analysis, analyze the relationship between image pixelsand cotton plant height. Establish the prediction model for cotton height prediction.8. Use system analysis method, analyze the overall structure of the cotton plantidentification system. Select the development environment and platform for the system.Design cotton plant identification system by the use of MATLAB R2011A GUI toolbox andM file.Aiming at the problems which existing in the process of cotton topping by machine, anew method for cotton plant identification based on machine vision technology was post inthis paper. The cotton plant identification system was build, the cotton identification, location,as well as the cotton plant height of the cotton plant measurement were achieved, which laidthe foundation for subsequent further research.