Online Detection of Abnormal Events in Video Streams

Table 2

AUC of abnormal event detection method of different features 𝐹 via original SVDD which learns training samples offline, Strategy 1 online one-class SVM, and Strategy 2 online one-class SVM. The biggest value of each method is shown in bold.

Features	Area under ROC
	Lawn	Indoor	Plaza
Training samples are learned offline
𝐹1(6×6𝑑𝑢)	0.9426	0.8351	0.9323
𝐹2(6×6𝑑𝑣)	0.9400	0.8358	0.9321
𝐹3(8×8)	0.9440	0.8375	0.9359
𝐹4(12×12)	0.9591	0.8440	0.9580
𝐹5(17×17)	0.9567	0.8649	0.9649
Strategy 1
𝐹1(6×6𝑑𝑢)	0.9399	0.8328	0.9343
𝐹2(6×6𝑑𝑣)	0.9390	0.8355	0.9366
𝐹3(8×8)	0.9418	0.8377	0.9411
𝐹4(12×12)	0.9581	0.8457	0.9573
𝐹5(17×17)	0.9551	0.8628	0.9632
Strategy 2
𝐹1(6×6𝑑𝑢)	0.9427	0.8237	0.9288
𝐹2(6×6𝑑𝑣)	0.9370	0.8241	0.9283
𝐹3(8×8)	0.9430	0.8274	0.9312
𝐹4(12×12)	0.9605	0.8331	0.9505
𝐹5(17×17)	0.9601	0.8495	0.9746