A procedure to compare the precision of random point and systematic grid point sampling methods for corn and soyabean residue measurement was presented. Both sampling methods were performed on images for a wide range of residue coverage and sampling points. Residue coverage of each image was measured using an automatic thresholding algorithm. A computer program carried out all the calculations for thresholding, sampling point generation, and point counting. Repealed estimations of residue cover in an image were normally distributed, and both sampling methods were unbiased estimators of residue cover. Systematic sampling was more precise than random sampling for both corn and soybean residue in most cases. Four coefficient of variation (CV) regression equations were developed. The CV equations of random sampling corresponded well to theoretical results. The CV equations of systematic sampling could be easily used to find the number of sampling points needed to achieve predefined precision. An example shows how to design a sampling pattern so that the least number of sampling points is used to make sure that there is 90% confidence that the estimated residue cover of an image is between 95 and 105% of real residue cover. This analysis procedure can be readily used to investigate other sampling methods and other types of residue.
|Original language||English (US)|
|Number of pages||9|
|Journal||Transactions of the American Society of Agricultural Engineers|
|State||Published - Sep 1 1995|