Soccer Ball Detection by Comparing Different Feature Extraction Methodologies

Table 1

Ball classification performance using SIFT and different number of clusters in the 𝑘-means algorithm on whole dataset.

SIFT descriptors length	TP	FN	TN	FP	Detection rate (%)
10	579	658	988	249	63.33%
20	621	616	1059	178	67.90%
30	558	579	1056	181	65.23%
40	519	718	1091	146	65.07%