THE SPATIAL LOW-PASS FILTERING AS AN ALTERNATIVE TO INTERPOLATION METHODS IN THE GENERATION OF COMBINE HARVESTER YIELD MAPS

摘要

Interpolation techniques are usually employed in the yield maps generation. In this article, a method based in spatial low-pass filtering is presented and compared with interpolation techniques. The method consists in the application of low-pass filters typically used for image processing. The yield map is constructed in a similar way to a Digital Elevation Model (DEM.), except for the fact that heights represent crop quantities. A combine harvester was equipped with a GPS receiver and a grain mass flow sensor. Two applications were written in Lab VIEW to carry out data acquisition, data processing, and yield map generation. The system was tested with a combine in a plot. Four yield maps were created as a raster image with a spatial resolution of 1 x 1m; two using spatial low-pass filtering with different filter masks, and two others using interpolation techniques. Interpolation techniques used were Inverse Distance Weighting (IDW) and Ordinary Kriging (OK). Due to the unfeasibility of applying a cross-validation process to the filtering-based method, an accuracy comparison could not be made. Thus, the quality of both methods was assessed according to five criteria: (i) the intrinsic robustness over ground speed and (ii) cutting width variations, (iii) the ability for preserving field boundaries and (iv) total crop mass, and finally (v) the processing time. Taking into account the implementation procedure presented and the results obtained, it can be concluded that implementation of a spatial low-pass filtering method is simple. It does not require speed nor cutting width information. It guarantees total grain mass conservation and requires low computational cost. Its use is advisable when speed and/or cutting width data are not available. The implementation of an interpolation-based method is more complex and more expensive. It also requires additional instruments to measure the machine speed. In the absence of a cutting width sensor, this method offers poor results when the real cutting width does not match the estimated cutting width. Interpolation methods do not guarantee total grain mass conservation. However, interpolation-based techniques allow error quantification and, thus, theoretically generate a more accurate yield map. This presented method could be implemented in special situations where a very simple system is needed, but very accurate maps are not necessary.