Research On Automatic Early Warning Robot Technology For Greenhouse Vegetable Diseases

With the improvement of people's living standards,greenhouse crops are becoming more and more common.Fresh and delicious fruits and vegetables can be eaten in different seasons,and more labor is needed for the growing area of planting.But human labor is time-consuming and laborious,and it is of great significance to improve agricultural modem planting and production technology.In this paper,the hardware and software parts of the greenhouse vegetable disease warning robot are studied.The hardware is mainly mechanical design and circuit connection,and the software part is mainly the advance control of the warning robot and the program control of machine vision.Image acquisition,processing and recognition based on machine vision are also studied.This paper mainly completed the following aspects of technical research:1)After consulting papers of the the greenhouse robot and plant disease early warning,the frame structure of the system design is determined.The mechanical and circuit parts of the forewarning robot are built.The unreasonable parts are improved,and the functions of each module are introduced.2)In this paper,the advance control scheme of the early warning robot is designed.We adopt the line patrol method based on machine vision.The moving route image is collected by the camera,the image is processed and recognized,and the moving line is detected by the Hough transform.The control parameters of stepping motor are extracted from the processing result,and the motion state of robot is adjusted in time.The image of moving route with leaf occlusion is processed,and the result is obvious.3)Taking the image of cucumber leaf powdery mildew disease as an example,the image was processed and recognized,and the theoretical basis was analyzed,including the conversion of color space RGB model to his model,image grayscale,denoising and threshold segmentation,edge detection and other algorithms.Several leaf image processing methods are compared,and a better performance image processing algorithm is selected.4)We use support vector machine classifier to detect and recognize the disease spots of cucumber powdery mildew,and extract the color and shape features of powdery mildew spots in cucumber leaves.Four kernel functions are used to train and detect the SVM classifier for the collected powdery mildew disease images.Compared with the results obtained by SVM classification training of these four functions,the SVM classifier based on radial basis function kernel function has better effect on classification of cucumber leaf powdery mildew.Through the test results,we can know that the accuracy rate of the greenhouse vegetable pest warning robot in the control process based on machine vision inspection is high,and the effect is obvious in the feature extraction and recognition of powdery mildew disease spot.