An Improvement of the Camshift Human Tracking Algorithm Based on Deep Learning and the Kalman Filter

Table 2

The experimental results of the proposed system and traditional Camshift.

Condition	Test case	Traditional Camshift [10]				The proposed system
		Number of frames	Number of false frames	Accuracy (%)	FPS	Number of frames	Number of false frames	Accuracy (%)	FPS
Good lighting room conditions (700 lux)	Multiple people using the same shirts	352	136	61.37	25	352	12	96.59	12
	Multiple people using the different shirts	405	37	90.86	25	405	9	97.77	12
Poor lighting room conditions (100 lux)	Multiple people using the same shirts	389	206	47.04	25	389	23	94.08	12
	Multiple people using the different shirts	481	131	72.77	25	481	25	94.80	11
Outdoor conditions (9522 lux)	Multiple people using the same shirts	326	118	63.80	25	326	10	96.93	12