Research On Precise Management Zones Extraction From Machine-Collected Cotton Fields

The establishment of accurate management zones cotton can effectively reduce the operating costs of cotton field and environmental pollution, as a result, achieving the sustainable use of cotton field. This study is to realize the accurate management zones of cotton field, which is one of the process of implementation of precision agriculture technology system, so import the Insight cotton yield monitor system of the AgLeader Company which is come from the U.S. Then, by using this equipment I got 2 years real-time data of cotton yield, and then analyzed and processed the errors of this data, also analyzed temporal and spatial variability in view of the processed data, and through the use of highly accurate interpolation method to generate spatial distribution map of cotton yield, ultimately, applied the k means clustering algorithm for establishing accurate management zoning map of cotton field. This study includes the following four aspects:(1) In order to accelerate the localization process of cotton yield monitoring system, detect the output of R & D working conditions of it in real-tine, but also to provide effective, reliable data to support for further optimization of monitoring system, researched and designed an indoor and online test bench of cotton yield.Put the the import of Insight for the cotton yield monitoring system in the interior line of cotton yield monitor test platform to examine, and the testing showed that it can be better to work in the Xinjiang cotton area of China, and provides a latter part of field test by using the the Insight cotton yield monitoring system with Reliability basis.(2) Installed the Insight cotton yield monitor system of the AgLeader Company, which is come from the U.S, on the John Deer 9970 cotton picker to do sustainable field experiments. According to this, on the one hand, further understood the composition of the cotton yield monitoring system, working process and the components working principle, setup and calibration methods; on the other hand, successfully obtained 2 years real-time data on cotton production. The results showed that cotton yield in the monitoring system must be properly installed and correctly set the required parameters; the correct height of the picking head, driving distance, the weight of cotton and cotton area under the premise of the calibration, only in this way, the yield data can be more realistic and reliable.(3)Through combination of field experiments and cotton harvest, a data filter was designed to weed out the errors from the cotton yield data, which contained the system errors, gross errors and random errors; comparing the establishment of cotton yield spatial distribution map around the processing errors, found that, after treatment, the singular points in cotton yield spatial distribution map were significantly reduced, the smoothness and cluster distribution were also markedly improved.(4)Analyzed temporal and spatial variability in view of the processed two years of cotton yield data, and it showed that two years of cotton yield data on the variation in temporal performance was not very strong, and in the spatial variation was on a medium intensity variation, which met the requirements of precision agriculture management zones and operations management variables. After that, used statistical analysis function of the ArcGIS software, adopted ordinary Kriging to select the best fitting model for the yield of high precision interpolation spatial distribution map.(5) In order to establish a two-year comprehensive spatial distribution map of cotton yield, Adopted normalized processing and interpolation to the 2 years real-time data of cotton yield,then,k means clustering algorithm was applied to establish the precise management zoning map of cotton field, and also establish management zoning map window by using different scales of smoothing filter to weed out spots or debris. Ultimately, determined that the most appropriate management of the number of partitions was 4, the most appropriate spatial scale of the filter was 18m.