Comparison of Two Techniques of Pattern Recognition in the Image Analysis-Based Wheat Stalk Length Characterization

The chaff quality or, more specifically, the distribution of stalk length after straw shredding during wheat harvest with a combined harvester is of significant interest for ploughless tillage practices. The currently applied characterization methods (manual length measurement or the cascade sieve analysis) are time-consuming and labour-intensive. Image analysis-based size characterization has the potential to solve these problems. In this study, two techniques of digital image processing, the well-known method of image moments (rectangular model) and a sub-pixel skeletonization approach (flow lines-tracing), were applied comparatively for stalk length measurement. Upon applying the rectangular model, the analyzed stalks were found to be longer than when the flow lines-tracing algorithm was applied. This was attributed to the unbiased decision criteria of the measuring method. A greater length of the stalk is to be expected in the main stalk than when the length of a twisted or branched stalk is measured from the edge length of a rectangular box that encloses it.