TABLE 3.

Classification training results of nitrogen application samples (n = 10 000).

Classifier type	Overall accuracy (%)	Wheat spike accuracy (%)	Background accuracy (%)
linSVM	86.40	87.30	85.31
quaSVM	88.49	90.27	86.30
cubSVM	83.33	76.38	90.27
finGSVM	88.49	88.29	89.28
medGSVM	89.78	91.27	88.30
coaGSVM	87.30	89.28	85.31
finKNN	86.70	86.30	86.30
medKNN	88.68	89.28	88.29
coaKNN	88.39	92.26	85.31
cosKNN	88.09	87.30	89.28
cubKNN	88.19	88.29	87.30
weiKNN	91.48	93.81	89.30