Research On Actuator For Laser Weeding Robot

This paper is a study in the field of farmland intelligent weed control, supported by national natural science fund project (60971115),"Modeling and Dynamic Simulation of Soil Compaction Effect on Plant Underground Tissue Architecture".Weeds is the most threat to agricultural production, and is also an important factor limiting agriculture production in our country.At present, the chemical weed control as the main way of weeding and the wide use of herbicides, although has obtained the remarkable result, but also brings many problems at the same time, straightforward type of energy waste caused by the spraying way cause resistant weeds, food safety problems caused by the pesticide residues, etc. For solving these problems, this study put forward weeding method based on the laser.This paper focuses on the field of laser weeding robot actuator is studied, the main research contents and results are as follows:1. Weeding technology at home and abroad and the research progress is summarized. This paper introduces the present situation of agricultural robots at home and abroad and weeding robot. The study status of weed identification and the principle of laser weeding technology and research progress is analysed.2. The design principle of the actuator to laser weeding robot is analyzed, mainly including:laser control part, weed identification part, robot localization and navigation field.3. The use of distance measurement based on image processing is a way that robot orientation in the field. By the theoretical analysis, the positioning method of benchmarking can accurately position the robot in the field. Test shows that the method based on image calibration ranging method is feasible and the error is low, can be used in practice. Through image processing in robot navigation, the navigation process need constantly obtain image information change the direction of the navigation in time.4. Dividing the image that obtain images of weed identification into100area and using the inverse operation of mechanical arm manipulator design which two motor vertical connection, the laser machine can put into each region and the corresponding motor running parameters of each region can be obtained. Optimizing the execute the order of the weed is done and the optimized result is more obvious.5. In the greenhouse, cultivation experiment was carried out, mainly for close to upland rice form of crops, upland rice, lawn grass and perennial ryegrass. Seven days after the emergence of the12days, data is measured, including leaf length, plant height, branch height and etc.. Images are obtained in20days, including41strains of upland rice,49strains of turfgrass and50strains of ryegrass, a total of140field image.6. The morphological characteristics of different plants were analyzed in different growth periods, and crop images were btained by applying the method of super green extraction based on Matlab7.8, the extraction result is ideal. Image information includes the R, G, B of image, image area, the minimum circumscribed rectangle parameters and circular degree. And using the parameters which obtained the information of plant to identify, the results show that the application of R different value, G different value, B different value, leaf width, leaf area, leaf perimeter and growth area are used as recognition parameters can more accurately identify three different plant.Based on laser weeding design study for the actuators of the robot, this paper mainly includes laser control part, weed identification part, the robot localization and navigation field. This paper proposed and implemented for upland rice recognition method of similar plants provides a new method for crop identification research; The field robot localization method based on image processing range provides a new way of thinking for the field localization; Vertical superposition method of motor control the laser.