Design On Machine Vision Navigation System Of Rice Transplanter And Path Tracking Experiment

When the rice transplanter works in the paddy field, it always run a straight line parallel to the transplanted rice seedlings and a "U" curve. Thus the driver is boring after a long time operation and the driver couldn’t concentrate on the driving, so the transplanting working quality will decrease. In addition, the speed and efficiency of the rice transplanter has been increasing in recent years, and the driver have to stop the transplanter frequently to add the rice seedlings to transplanter, which decrease the efficiency of the high performance transplanter. For decreasing the intensity of labour of the driver and to be free the driver to do some other operations such as adding rice seedling, it is necessary to develop a mechanism to steer the steering wheels automatically or semi-automaticlly. A navigation system of the rice transplanter based on machine vision will be developed to steer the steering wheel semi-automatically in this research, these contents include:Design on the automatic steering system of the rice transplanter. With the Kubota NSD-8 rice transplanter as study platform, on condition that guarantee the usability of existing steering mechanism, a kind of automatic steering system was designed. The mechanical system mainly comprised steering mechanism, clutch control mechanism, velocity control mechanism, lifting control mechanism, and the electrical part mainly comprised microcomputer control unit, remote transmitter installation and receiver installation, angle sensor, stepping motor. It was tested about 180° and 90° steering operation on the flat ground, the results showed that automatic steering system can steer the rice transplanter to achieve predetermined operation in remote control status.Design on the navigation system of the rice trasplanter. The camera and lens are selected after calculation, The size of field of view is that the width is 2.8 m and 1.2 m respectively, and the longth is 2.5 m. the height of the camera is 1.1 m. the camera model is Suntime 300C, the lense is Kowa 4.4-11 mm. The USB interface is used to connect the camera and the computer to construct the hardware of the navigation system. The navigation system software is programmed with MATLAB in the Windows based on the VFW technology. The test result shows that the hardware and software is able to get the image of the paddy field and can be used in the research.Research on the directrix recognition algorithm of vision navigation rice transplanter. R,G,B pixel values of rice seeding, residual straws, foam, water, mud are read manually. Color components were drawn in RGB, HIS, Ⅰ1Ⅰ2Ⅰ3 color space gray histograms. For the difference, Ⅰ3 color component gray value was used to segment the paddy field image. The rice seedling pixels were segmented from background pixels with threshold calculated with OTSU to get the binary image. A mask was put on the binary image to delete the left and right part of the upper image of rice seeding to get a better recognition precision. The location points in the image were detected with vertical projection mothod. Mean value of the difference of the location points were used to detect the fake location points and the discontinuous rows. Nearest Neighbor method was used to classify the location points. Directrixs were fitted with the Robust Regression. The results showed that minimum average error rate of one image was 0.78%, the average error rate of 20 images were 2.33%, and the rice transplanter was feasible to get the navigation parameters with the algorithm.Research on the algorithm of the navigation information acquisition of the rice transplanter. The mapping relation of the feature points of the rice seedlings in image coordinates and world coordinates was analyzed. The pose relation model of the rice seedlings directrix in image coordinates and world coordinates was constructed with the pin-hole camera model and the Hough Transformation. The navigation parameters in the single frame are acquired with Hough Transformation, and converted into World Coordinate System with coordinate transformation to get the navigation parameters in the World Coordinate System. The artificial lane was paved to imitate the rice seedlings, and the single frame navigation information acquisition experiment was carried out. The results showed that the mean error of the head angle was 4.76°, the variance of the heading angle was 5.07°, the mean error of the position offset was 4.95mm, the variance of the position offset was 5.42mm. static experiment on single frame navigation parameter acquisition results showed that the navigation parameters would fluctuate in a certain range influenced by the vibration of engine and variation of the light condition.Based on the continuity of the images in the contious frames of video, average filtering method is used to smooth the navigation parameters in the contious frames, experiment on navigation parameters in multi-frame acquisition results showed that this method will make the variation range decrease to a acceptale level.Design on the navigation controller of the navigation system of the rice transplanter. Based on the driver simulation technology, heading angle and position offset are inputs of the controller, steering angle of the steering wheel is the output of the controller, then a two-dimension fuzzy controller was designed. The simulation results of rice transplanter path tracking to ramp response step response of 30 cm is analyzed, and the steering operation rules of rice transplanter path tracking is concluded. The path tracking experiment results of rice transplanter driven manually showed that the steering rules of the rice transplanter is same to the steering rules generating by the fuzzy controller. It means that the fuzzy controller in the research will be able to navigate the rice transplanter.Experiment on vision navigation path tracing of the rice transplanter. When the rice transplanter is driven manually, the peak value distance of path is 4.2m, and the adjustment distance 4.6m, steady-state error is 0.005m. when the transplanter in the pose that heading angle is 0°nd the position offset is 0.3m, and the transplanter is driven by the navigation system to track a certain path, the peak value distance of path is 3.6 m, and the adjustment distance 5.4 m, steady-state error is 0.01m. The field experiment result of the path tracking in paddy field of vision navigation rice transplanter is that the maximum deviation is 24cm, the average 7.9cm. The experiment results showed that the navigation system is able to guide the rice transplanter to track the transplanted seedlings and we achieve our expected aim.