Design And Experiment Of Intelligent Intra-row Mechanical Weeding Device For Lettuce Based On Vision

Weeds can be divided into inter-row weeds and intra-row weeds according to their growing positions.Compared with inter-row weeds,intra-row weeds are closer to crops,and the intra-row weeding area is discontinuous,which makes mechanical inter-plant weeding difficult,so intra-row mechanical weeding technology has been slow to development.As the application of vision and machine learning technology in agricultural machinery and equipment,the difficulty of weeding between plants has been greatly reduced.At present,the recognition of crops and weeds is only a single extraction of color,shape and texture.After the feature extraction is completed,the support vector machine and other classifiers are used to classify them.In the real scene,because of the different shapes of weeds,some weeds and crops will have some problems such as mutual shading of leaves,which makes it difficult for classifiers such as support vector machines to meet the classification requirements.In view of the shortcomings of the current intra-row mechanical weeding device and seedling identification methods,this paper designs an intelligent intra-row weeding device.The device composed of a cam combined intra-row weeding mechanism and the seedling grass recognition and segmentation model to cooperate with each other to achieve weeding.The main research contents are as follows :(1)A cam combined intra-row weeding mechanism is designed.Through the cam combined transmission mechanism,the weeding knife swings in the given motion law,so as to realize intra-row weeding and avoid seedling lettuce in time.Firstly,the mathematical model of the motion trajectory of the intra-row weeding mechanism is established.According to the mathematical model,the optimization program of the cam combined intra-row weeding mechanism is compiled by MATLAB,and then a set of design parameters that meet the weeding requirements are finally selected through human-computer interaction.Then the mechanical model of the weeding knife is established,and the spring parameters satisfying the conditions are obtained according to the mechanical model.Finally,the three-dimensional modeling of the cam combined intra-row weeding mechanism is carried out,and the three-dimensional model is imported into ADAMS for simulation analysis.By comparing the simulation weeding trajectory with the theoretical weeding trajectory,it is found that the two are basically the same,which proves the correctness of the mathematical model of the intra-row weeding mechanism.(2)A method to identify seedling lettuce and weeds that based on an image block and support vector machine is proposed.It realizes their precise identification and boundary segmentation.Firstly,the Otsu algorithm is used to binarize the images,and the region labeling algorithm is used to mark the region where the green plants are located.The region(region of interests,ROIs)completed by each marker is normalized to 256 × 256 pixels.At the same time,the image block technology is introduced to block each ROIs,and the three texture features of HOG,LBP and GLCM of each image block are extracted.GA-SVM,SVM,RF and KNN are used to classify different feature fusion strategies with block size of 16 × 16 pixels,32 × 32 pixels and 64 × 64 pixels respectively.The results show that when the block size is 32 × 32 pixels and the data set of LBP and GLCM feature fusion is used,GA-SVM can achieve the highest accuracy.Aiming at the problem that GA-SVM classification model also has error recognition,an image block reconstruction method based on the comparison of the center point and eight-neighbor label value is proposed,and this is combined with the proportion of image blocks of two labels for comprehensive judgment,thereby increasing the overall recognition accuracy.(3)According to the two-dimensional drawings,the cam combined intra-row weeding mechanism was processed,and the prototype is built.The completed prototype is installed on the mobile chassis and combined with the seedling identification system.According to the test conditions,the required indoor soil ridge was designed,and the cam combined intra-row weeding device was debugged.The weeding device is tested in the indoor soil ridge,and the problems of injured seedlings and no weeding in the test were analyzed in detail.The test results can also prove that the developed intelligent intra-row weeding device can meet the needs of intra-row weeding operations.