Research Of Path Identification Of Small Agricultural AGV In Corn Based On Machine Vision

With the development of computer technology and automatics rapidly,s were applied in industry,medicine,military,agriculture and other fields.China is a great agricultural country,so it is widely concerned to improve the efficiency of agricultural production and the safety of operations of agricultural.Agricultural robots are the important parts of modern agriculture and precision agriculture,and are also the direction of agricultural development in the future.The current researches are concentrated on installing related sensors such as GPS,IMU,speed sensor and camera on existed or newly designed platforms of agricultural machinery as elements of hardware acquisition.The current researches are concentrated on installing related sensors such as GPS,IMU,speed sensor and camera on existed or newly designed platforms of agricultural machinery as elements of hardware acquisition.Based on the navigation path of identified and planned,IPC analyzes and makes decisions.Under receiving commands,controllers control the machine to steer and walk in accordance with the identified navigation path.This research was based on existing navigation path recognition algorithms of agricultural robots,and a method of path planning of a small agricultural AGV was proposed based on monocular vision considering the treatment of diseases and insect pests of middle and later period's corns and the demand of autonomous navigation of precise agriculture.The specific contents as follows:First of all,The model machine and the control system of agricultural AGV was designed.Agricultural AGV was wheel style,so that it had light weight and strong flexibility.Considering the pattern of corn planting in North China,the size of model machine was designed and the spare for thermal fogger and spraying system was left.Agriculture AGV was drived by rear-wheel using a motor.The steering engine controlled steering of front wheel.Laptop and image processing board were used to process and analyze images.Secondly,the weight of three components in traditional 2G-R-B algorithm was improved in the RGB color space.Some images of corn field were collected and green features in each image were extracted respectively,then the average of three components of G value was calculated and the coefficient of green feature extraction separating crop rows.The binarization method was used to preprocess image,then the opening operation and closing operation,corrosion and expansion were combined to segment the ROI(Region of Interest,ROI).In the aspect of identification and extraction of loaction points of corn roots,different methods were compared and a peak point detection algorithm based on vertical projection drawing was proposed to judge the peak points of corn roots.The precise localization points of corn roots were obtained after second-order correction.Finally,three kinds of linear fitting methods were used to fit localization points respectively.The results showed that the least square method can be faster and more accurate to fit the corn root lines.The angle bisector of two root lines was calculated and used as navigation line of agricultural AGV.The navigation angle and lateral deviation of centerline of corn rows were inputs of navigation to control walking and steering of AGV after image coordinate system was converted to world coordinate system by calibration of HDL and camera.The experiments showed that the proposed method of this paper applied to pests control of high stalk crops such as middle and late corns.And it satisfied the needs of vision navigation of small agricultural AGV.The average of lateral deviation was 3 cm and angle of navigation deviation average was 2.06 °.So it can satisfied the needs of field navigation.This paper provided a reliable reference of autonomous navigation for agricultural AGV in corn field.