Robust Grape Detector Based on SVMs and HOG Features

Table 6

Evaluation of the detectors trained on T-3, set-up according to Table 3, on original datasets of grape type (G), where and . Following versions of the detectors evaluated using the measures (1a), (1b), and (1c): is detector without any conversion, is detector with the conversion according to ITU-R recommendation BT.601, and R is detector with the generalized conversion (2).

Kernel	Version	G-1			G-2			G-3			G-4			G-5
		Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall
Linear		86.25%	0.9560	0.7600	81.00%	0.9697	0.6400	87.00%	0.9933	0.7450	85.75%	0.9673	0.7400	84.75%	0.9728	0.7150
		86.25%	0.9618	0.7550	79.25%	0.9606	0.6100	86.00%	0.9865	0.7300	87.50%	0.9870	0.7600	84.75%	0.9793	0.7100
		87.75%	0.9576	0.7900	80.25%	0.9618	0.6300	86.50%	0.9803	0.7450	86.75%	0.9565	0.7700	86.00%	0.9500	0.7600
RBF		88.75%	0.9874	0.7850	81.00%	0.9844	0.6300	89.00%	1.0000	0.7800	87.25%	0.9934	0.7500	88.00%	0.9935	0.7650
		89.50%	0.9877	0.8000	81.00%	0.9844	0.6300	88.25%	1.0000	0.7650	87.25%	0.9869	0.7550	87.75%	0.9935	0.7600
		89.00%	0.9699	0.8050	82.25%	0.9924	0.6500	87.00%	0.9933	0.7450	87.25%	0.9806	0.7600	87.75%	0.9935	0.7600