Alert volume normalization in a video surveillance system
View Patent ↗Techniques are disclosed for normalizing and publishing alerts using a behavioral recognition-based video surveillance system configured with an alert normalization module. Certain embodiments allow a user of the behavioral recognition system to provide the normalization module with a set of relative weights for alert types and a maximum publication value. Using these values, the normalization module evaluates an alert and determines whether its rareness value exceed a threshold. Upon determining that the alert exceeds the threshold, the module normalizes and publishes the alert.
1. A method for normalizing and publishing alerts generated by a behavioral recognition system evaluating a sequence of video frames, the method comprising:
receiving an alert having a type and an original rareness value, wherein the alert corresponds to an observed behavior of one or more foreground objects detected in the sequence of video frames and wherein the original rareness value reflects how common the alert is based on alerts of the type that have previously been generated by the behavioral recognition system while evaluating the sequence of video frames;
converting the original rareness value to an alert percentile value;
upon determining that the alert percentile value is greater than a percentile threshold, normalizing the alert, wherein the percentile threshold is
ξ
i
=
1
-
v
i
N
i
,
where v i is a maximum allowed alert counts for the alert type value and N i is an estimated next-day alert counts value; and
publishing the alert.
2. The method of claim 1 , wherein the percentile threshold is
ξ
i
=
N
i
+
p
i
-
(
1
+
λ
)
v
i
N
i
if
p
i
>
v
i
,
where p i is a number of published alerts for a current day and λ is an overage value.
3. The method of claim 1 , wherein normalizing the alert further comprises calculating a normalized rareness value
η
i
=
ɛ
i
-
ξ
i
1
-
ξ
i
,
where ε i is the alert percentile value and ξ i is the percentile threshold.
4. The method of claim 3 , wherein the normalized rareness value is
η
i
=
1
-
rand
m
+
∑
k
=
1
B
m
k
p
+
∑
k
=
1
B
p
k
if
r
=
1
,
where m is a current number of alerts having an original rareness value being 1 for a current day, p is a current number of published alerts for a current day, and is the original rareness value.
5. The method of claim 1 , wherein publishing the alert further comprises calculating a publication rank value β=min(P, N p )*(1−η i ), where P is a maximum publication value, N p is a rank-renormalization constant, and η i is a normalization rareness value.
6. The method of claim 5 , further comprising upon determining that the publication rank value is less than a maximum dispatch value, dispatching the alert.
7. A computer-readable storage medium storing instructions, which, when executed on a processor, performs an operation for normalizing and publishing an alert generated by a behavioral recognition system evaluating a sequence of video frames, the operation comprising:
receiving an alert having a type and an original rareness value, wherein the alert corresponds to an observed behavior of one or more foreground objects detected in the sequence of video frames and wherein the original rareness value reflects how common the alert is based on alerts of the type that have previously been generated by the behavioral recognition system while evaluating the sequence of video frames;
converting the original rareness value to an alert percentile value;
upon determining that the alert percentile value is greater than a percentile threshold, normalizing the alert, wherein the percentile threshold is
ξ
i
=
1
-
v
i
N
i
,
where v i is a maximum allowed alert counts for the alert type value and N i is an estimated next-day alert counts value; and
publishing the alert.
8. The computer-readable storage medium of claim 7 ,
wherein the percentile threshold is
ξ
i
=
N
i
+
p
i
-
(
1
+
λ
)
v
i
N
i
if
p
i
>
v
i
,
where p i is a number of published alerts for a current day and λ is an overage value.
9. The computer-readable storage medium of claim 7 , wherein normalizing the alert further comprises calculating a normalized rareness value
η
i
=
ɛ
i
-
ξ
i
1
-
ξ
i
,
where ε i is the alert percentile value and ξ i is the percentile threshold.
10. The computer-readable storage medium of claim 9 , wherein the normalized rareness value is
η
i
=
1
-
rand
m
+
∑
k
=
1
B
m
k
p
+
∑
k
=
1
B
p
k
if
r
=
1
,
where m is a current number of alerts having an original rareness value being 1 for a current day, p is a current number of published alerts for a current day, and r is the original rareness value.
11. The computer-readable storage medium of claim 7 , wherein publishing the alert further comprises calculating a publication rank value β=min(P, N p )*(1−η i ), where is a maximum publication value, N p is a rank-renormalization constant, and η i is a normalization rareness value.
12. The computer-readable storage medium of claim 11 , further comprising upon determining that the publication rank value is less than a maximum dispatch value, dispatching the alert.
13. A system comprising:
a processor; and
a memory for hosting an application, which, when executed on the processor, performs an operation for normalizing and publishing an alert generated by a behavioral recognition system evaluating a sequence of video frames, the operation comprising:
receiving an alert having a type and an original rareness value, wherein the alert corresponds to an observed behavior of one or more foreground objects detected in the sequence of video frames and wherein the original rareness value reflects how common the alert is based on alerts of the type that have previously been generated by the behavioral recognition system while evaluating the sequence of video frames,
converting the original rareness value to an alert percentile value,
upon determining that the alert percentile value is greater than a percentile threshold, normalizing the alert, wherein the percentile threshold is
ξ
i
=
1
-
v
i
N
i
,
where v i is a maximum allowed alert counts for the alert type value and N i is an estimated next-day alert counts value, and
publishing the alert.
14. The system of claim 13 , wherein the percentile threshold is
ξ
i
=
N
i
+
p
i
-
(
1
+
λ
)
v
i
N
i
if
p
i
>
v
i
,
where p i is a number of published alerts for a current day and λ is an overage value.
15. The system of claim 13 , wherein normalizing the alert further comprises calculating a normalized rareness value
η
i
=
ɛ
i
-
ξ
i
1
-
ξ
i
,
where ε i is the alert percentile value and ξ i is the percentile threshold.
16. The system of claim 15 , wherein the normalized rareness value is
η
i
=
1
-
rand
m
+
∑
k
=
1
B
m
k
p
+
∑
k
=
1
B
p
k
if
r
=
1
,
where m is a current number of alerts having an original rareness value being 1 for a current day, p is a current number of published alerts for a current day, and r is the original rareness value.
17. The system of claim 13 , wherein publishing the alert further comprises calculating a publication rank value β=min(P, N p )*(1−η i ), where P is a maximum publication value, N p is a rank-renormalization constant, and η i is a normalization rareness value.
18. The system of claim 17 , further comprising upon determining that the publication rank value is less than a maximum dispatch value, dispatching the alert.