| With the development of global information technology, digital agriculture has become the trend of modern agriculture. Through the digital design, visual expression and intelligent control all along the agriculture process (pre-, during and post-production), a timely adjusted chain that realizes the precision variety selection, precision planting, precision fertilizer, precision irrigation, grading storage and dynamic monitoring can be formed to achieve the maximum input-output ratio, which has a very important significance to the realization of agricultural sustainable development.In this study, embedded network technology and remote field information monitoring system served by CDMA were used as the basic tools to obtain the real-time image of winter wheat and summer maize, and the remote automatic wireless transmission of information in a farm of North China Plain. The remote dynamic monitoring of the height of winter wheat and summer maize was realize by the means of digital image processing technology, which can provide technical support for the precision management and digital management of the winter wheat and summer maize. And then it can provide decision basis to the precision management of winter wheat and summer maize in the North China Plain.Programming by Visual C++6.0, using the basic principles of "binocular stereo vision" to deal with dynamic binocular image pairs of winter wheat and summer maize transferred to the terminal server, getting the correspondence of target point in the left-right image through camera calibration, image segmentation, noise reduction, regional grid region matching process and so on, building the three-dimensional information of objects which is similar to three-dimensional perception process of human visual system, by which the real-time height of winter wheat and summer maize can be obtained quickly. The major results are as follows:(1)The method of calibration bar image was used to calibrate the camera, and four calibration bars were inserted in the camera monitoring area at the four corners of the One square meter square.The calibration bars were painted in red, pasted white calibration points at the top, middle and bottom, with a total of 12 points. When the calibration bars were set up, the cameras were adjust the left-right to make all the calibrations fall into the image and made sure the size of the left-right images are roughly the same using the calibration bar as the target to focus on. When they were ready, the left-right image pairs were collected immediately, using the linear method to calibrate the camera. The study found that the standard deviation of the calibration errors is in the allowable range, as a fast and effective camera calibration method.(2) For the images of winter wheat and summer maize in mid-late period, after the grayed the images using the G component, the discrimination of crops was well by finding the threshold automatically of biggest intra-class variance. For the image of winter wheat and summer maize in the early stage, using of the G component gray can hardly discriminate the objects.The study found that, for the images of winter wheat and summer maize less affected by light or collected in the cloudy days, using the weighted value (2G-R-B), and automatically threshold finding of biggest intra-class variance can also find the target partition. However, for the seedling images of winter wheat and summer maize affected strongly by the light, a good segmentation method has not found now.(3) For the solving of the problems that the crops in intercropping system were disorganized, shading one another and resulted to poor image resolution, we try to match the grid of the left-right projection region. First, through the point in the calibration bar obscured by crops, the height of projection area can be confirmed preliminary.Second, the projection areas in the images were girded and then matched. Only the interested white part of the images was matched, and the totally white pixels of the grids were matched directly. For the partly white pixels of the grids, a matching criterion was set up. When a grid of white pixels accounts for the proportion of total pixels within the grid reaches a certain value, the grid can complete the match successfully, or else it will be considered as the background. Then the mass center of the successfully matching grid was extracted and rebuild three-dimensionally. The Y value of the centroid point was considered as the height of the plant.(4) Based on the method mentioned above, images of winter wheat and summer maize were analyzed, and the result of the image analysis reached a certain degree of accuracy. The relativity of the measured height and the corresponding image analysis height of the plant reached 0.9129 and 0.9848 respectively. The best correlation of the image analysis of winter wheat and summer maize and the measured results were displayed in the 30 (0.1)and 10 (0.1)combinations.Fitting equation of measured value and image analysis result were:y=1.1178x+6.2503;y=0.9585x+50.857.In the model checking, the maximum relative error of winter wheat is 8.37%, the minimum is 2.58%; the maximum relative error of summer maize is 8.84%, the minimum is 0.11%.From the experiment result, we can draw the conclusion that the idea of the plant height analysis in the test is feasible, and the analysis reaches a certain degree of accuracy. During the process of experiment, the number of regional grid can not be too few for winter wheat; the number of regional grid maintains at about 100 for summer maize, which not only improve the accuracy of the image analysis, but also save the computer calculation.(5) The remote real-time dynamic monitoring system of growing situation for winter wheat and summer maize was carried out by using the MFC program of Visaul C ++6.0.Modular design ideal was applied during the development process, and the main task of the software was encapsulated in a few major classes.A timer and threading were used to achieve the real-time image searching, dynamic preprocessing, automatic analysis, the dynamic displaying of results.Then the three-dimensional display of the matching point can be achieved using the OpenGL window. |
PDF Full Text Request