Research On Spatial Visual Perception And Picking Control Method Of Picking Robot In Agricultural Environment

In this paper,fruit perception,plucking position space perception and intelligent control method of picking robot were studied to achieve the goal of fruit picking.There is absolute academic research value for the development of agricultural modernization to realize the picking robot’s intelligence.Specific research contents include:⑴ Research Fruit Perception.A fruit perception method based on wavelet transform and Otsu threshold denoising is proposed,which solves the low fruit perception effect caused by the error or absence of fruit feature extraction information due to illumination and cyan skin in agricultural dynamic unstructured environment.Firstly,the color model of fruit image is transformed,and the color model with the fruit feature is selected.Secondly,wavelet transform method is introduced to extract fruit features with appropriate color components image,and wavelet transform sub-images containing main features are selected.Thirdly,FCM algorithm is used to segment the wavelet transform sub-image;Finally,the image denoising method based on Otsu threshold is adopted to solve the problem that noise interference still exists in the segmented target area.The perception accuracy of cyan and orange were 87.10% and 94.18%,respectively.The perception accuracy of navel orange is 92.96% and 90.15% respectively under light and backlight,and90.82% and 93.18% respectively under occluded and unoccluded.The total perception accuracy was 92.07%.The method has strong environmental adaptability and is suitable for the identification and processing of navel orange images under different occlusion in agricultural environments.⑵ Research Fruit Plucking Position Perception.A bionic motion-based plucking position perception method is proposed,which solves the problem that spatial location information of fruits is vulnerable to noise caused by factors in the agricultural dynamic unstructured environment,such as light and angle,etc.Firstly,the coordinate information of fruit in two-dimensional image is determined by CFCM(Circle filling method based on the centre of mass)method;Secondly,the three-dimensional spatial coordinates of fruit are determined by coordinate transformation;Finally,the plucking position perception based on bionic movement is realized by controlling the rotation angle signals of each picking robot’s mechanisms.Within the allowable range of displacement error and standard deviation,the plucking position perception method has a better picking accuracy of 95%and a faster perception time of 0.33 s.⑶ Aiming at the demand of fruit picking in agricultural dynamic unstructured environment,a picking control method based on Fmincon and PD(Proportional–Differential)angle signal is proposed.Firstly,the Solidworks,Adams and MATLAB/Simulink toolbox is used to build the picking robot’s simulation model platform.Secondly,the three-dimensional simulation model and it’s mathematical analysis model are established.Thirdly,the Fmincon method in MATLAB toolbox is used to analyze the nonlinear equation with the angle value as the control quantity,and get the optimal rotation angle of each picking robot’s mechanism.Finally,the signal control method of picking mechanism rotation angle based on optimized PD is designed.The mean displacement error of the straight-line distance between the measuring point at the end of the picking mechanism and the target point is 6.00608 mm,and the standard deviation is 1.0408286.It realizes the autonomous fruit’s picking of by the picking robot in the dynamic unstructured environment of agriculture.