Determinação e remoção de erros em mapas de colheita de milho

Abstract

The automation of the the monitorated harvest ends introducing systematic errors to the process, for this reason, a methodology of filtering of data became necessary for obtaining confident maps and it treats about the mean error factor, which is the delay time. The delay time is that spent between the cutting of the plant by the platform s bar and the passage of the grain by the sensor of productivity and humidity. For its determination, a data set will be analysed with 10 different delay time, from 0 to 27 seconds, in an interval of 3 seconds. The filtering begins by the determination of the time of stuffing and emptying that is the one which all the spaces and internal components of the combine are filled or deflated by the grains mass occurred in each input and output of the combine. In this process were eliminated 5,69% of the total points, being the process more efficient in the elimination of points, because it acts mainly in the plot s heads, where the most errors occurs. For the width error of platform were eliminated only 0,09% of the points of data. At the points having sign fails of GPS differential correction, in average 2,04% were eliminated. In the filtering of data, in average 2,27 % of the points were eliminated. In localization errors, all the points having until one meter of distance were removed, eliminating in average 0,14% of the data. At the end of the filtering process, the subsequent phase was the determination of the correct delay time for the variables humid productivity, humidity and elevation. As the delay time is directly linked with the combine speed inside the transport, areas with homogeneous speed were considerated. Each one of the 12 sub- areas from 10 data sets were submitted to the process of analysis of spatial dependence of the better semivariogram´s adjust. The trend more emphasized in the tabulation of the semivariogram parameters is the linearity of the nugget effect for the factor elevation, originating from own absence of the delay time of this data, since the elevation is passed joinly to the geodetic coordinates by DGPS, for the harvest monitor. The determination method of the delay time and the filtering process were satisfactory, meanly for humid productivity and humidity, despite of the high natural data variability.