Traffic Flow Parameters Collection under Variable Illumination Based on Data Fusion

Table 2

Results of detection performance evaluation on the testing sets. On average of four testing subsets, the mean values of [Precision, F-measure] obtained by different methods are as follows: MOG [76.94%, 14.93%], MultiLayer [78.40%, 71.29%], CenterNet [86.17%, 89.97%], proposed method [88.03%, 91.10%].

Daytime-testing 1
Method	Precision	Recall	F-measure	NFP/image	NFN/image	mAP
MOG	83.27%	9.83%	16.76%	0.13	47.21	74.86%
MultiLayer	85.66%	78.29%	77.42%	6.59	15.62	79.93%
CenterNet	96.32%	92.75%	94.36%	0.95	2.79	97.57%
Proposed	97.37%	93.26%	93.21%	0.83	2.46	98.1%
Daytime-testing 2
MOG	86.18%	9.33%	17.14%	0.55	50.73	73.24%
MultiLayer	85.37%	69.86%	74.36%	7.48	17.23	87.4%
CenterNet	94.79%	91.65%	97.6%	0.58	2.49	98.2%
Proposed	96.3%	94.34%	96.83%	0.73	2.52	97.93%
Nighttime-testing 1
MOG	73.49%	7.28%	11.45%	1.90	43.69	58.42%
MultiLayer	74.28%	61.16%	62.13%	3.51	2.24	54.03%
CenterNet	77.21%	87.48%	83.93%	2.23	0.85	75.99%
Proposed	79.10%	95.44%	86.82%	1.89	0.75	88.39%
Nighttime-testing 2
MOG	64.81%	6.72%	14.36%	2.13	50.73	52.56%
MultiLayer	68.27%	64.62%	71.26%	2.87	2.62	51.78%
CenterNet	76.36%	89.14%	83.97%	1.81	1.10	75.37
Proposed	79.33%	93.12%	87.54%	1.34	1.82	86.27%