KR101480636B1 - Monitoring method of video monitoring device - Google Patents

Abstract

The present invention relates to a method of monitoring a video monitoring device, which is capable of automatically learning a target to be monitored with respect to each monitoring area. The method includes a monitoring region setting step of setting a monitoring region to be photographed by a pan-tilt-zoom (PTZ) camera; a first extraction step of extracting data of all objects moving as the PTZ camera photographs the monitoring region; a monitoring target object selection step of selecting a desirable object to be monitored in the monitoring region by a control unit; a first learning step of learning a classifier through classified data as data of the object to the monitored are classified in data of the object extracted in the first extraction step; a second extraction step of extracting data of all objects moving by photographing the monitoring region by the PTZ camera; a second learning step of inputting the data of the object extracted in the second extraction step into the classifier learned in the first learning step to classify of the data of the object to be monitored and learning the classifier based on the classified data. If an amount of data of the object to be monitored is greater than a reference value, the monitoring of the object to be monitored is initiated.

Description

[0001] The present invention relates to a monitoring method of a video monitoring apparatus,

The present invention relates to a monitoring method of a video surveillance apparatus.

In general, surveillance systems through cameras require monitoring personnel to detect specific objects. However, the monitoring agent misses many events over time. Of course, an intelligent surveillance system using a typical camera can automatically detect a specific object, but the surveillance area is limited because the camera is fixed. Therefore, an intelligent system capable of automatically detecting a specific object for each surveillance area is needed.

Korean Patent Laid-Open No. 10-2012-0068150 (June 27, 2012)

The present invention provides a surveillance method of a video surveillance apparatus capable of automatically learning a surveillance target for each surveillance area.

A surveillance method of a video surveillance apparatus according to the present invention includes a surveillance area setting step of setting a surveillance area to be photographed by a PTZ camera; A first extraction step in which the PTZ camera captures a surveillance area and extracts data of all moving objects; An object to be monitored to select an object to be monitored in the surveillance region; A first learning step of classifying data of an object of observation among data of the object extracted in the first extraction step and learning the classifier through the classified data; A second extracting step of re-extracting data of all moving objects by re-capturing the surveillance area by the PTZ camera; And a second learning step of inputting data of the object extracted in the second extracting step to the classifier learned in the first learning step to reclassify the data of the object to be monitored and re-learning the classifier through the reclassified data, Wherein monitoring of the monitored object is started when the data amount of the monitored object is larger than the reference value.

In the first learning step, the data of the object to be monitored may be classified into the selection data, and the data of the object other than the object to be monitored may be classified as the non-selection data.

In the first learning step, when the selection data is input to the classifier, 1 is output, and when the non-selection data is input to the classifier, the classifier can be instructed to output -1.

In the second learning step, when the data of the object extracted in the second extraction step is input to the classifier, if the output of the classifier is larger than 0, it can be reclassified as selection data.

Further, in the second learning step, when the selection data is input to the classifier, the classifier can be re-learned so as to output 1.

In the second learning step, when the data of the object extracted in the second extraction step is input to the classifier, if the output of the classifier is less than 0, it can be reclassified as non-selection data.

In addition, in the second learning step, the classifier can be re-learned to output -1 when the non-selection data is input to the classifier.

Also, if the amount of data of the object is smaller than the reference value, the second extraction step may be returned to the second extraction step.

In the monitoring area setting step, a plurality of monitoring areas may be set, and in the monitoring object selecting step, different monitoring objects may be selected for each monitoring area, and multiple monitoring objects may be simultaneously monitored.

The surveillance method of the video surveillance apparatus according to an embodiment of the present invention can automatically learn objects to be watched by performing class teacher learning through a classifier.

In addition, the monitoring method of the video surveillance apparatus according to an embodiment of the present invention can automatically monitor multiple objects for each surveillance area by using a PTZ camera.

1 is a block diagram illustrating a video surveillance apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a monitoring method of a video surveillance apparatus according to an embodiment of the present invention.
3 is a detailed flowchart illustrating a monitoring method of a video surveillance apparatus according to an embodiment of the present invention.
4A to 4C are video images for explaining a monitoring method of the video surveillance apparatus shown in FIG.

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the present invention.

1 is a block diagram illustrating a video surveillance apparatus according to an embodiment of the present invention.

Referring to FIG. 1, an image surveillance apparatus 100 according to an exemplary embodiment of the present invention includes a PTZ camera 110 for photographing a surveillance region, a storage unit for storing images and data captured by the PTZ camera 110 A controller 120 for controlling the PTZ camera 110, the storage 120 and the classifier 130 to classify data stored in the storage unit 120, 140). Hereinafter, a method of monitoring the object to be sensed by the video surveillance apparatus 100 will be described in detail.

2 is a flowchart illustrating a monitoring method of a video surveillance apparatus according to an embodiment of the present invention. 3 is a detailed flowchart illustrating a monitoring method of a video surveillance apparatus according to an embodiment of the present invention. 4A to 4C are video images for explaining a monitoring method of the video surveillance apparatus shown in FIG.

Referring to FIG. 2, a monitoring method of a video surveillance apparatus according to an exemplary embodiment of the present invention includes a monitoring area setting step S10, a first extracting step S20, a monitoring object selecting step S30, (S40), a second extraction step (S50), a second learning step (S60), and a surveillance determination step (S70). Hereinafter, the respective steps of FIG. 2 will be described with reference to FIGS. 3 to 4C.

The monitoring area setting step S10 is a step of setting a monitoring area to be monitored by the PTZ camera 110. [ First, as shown in FIG. 4A, a PTZ camera is installed on a ceiling, an outer wall, or a column of a surveillance area, and is operated by a PAN operation rotated in a horizontal direction and a tilt operation rotated in a vertical direction, The surveillance area can be easily photographed with the camera. In addition, the PTZ camera 110 also includes a zoom function, so that a specific area can be enlarged and photographed.

In the monitoring area setting step (S10), the controller 140 may set a plurality of areas to be monitored. As shown in FIG. 4B, the controller 140 may divide the surveillance area to be monitored by the PTZ camera 110 into areas a to l. Of course, the controller 140 can distinguish more or less.

The first extracting step S20 extracts all objects moving in the monitoring area. In the first extraction step S20, the PTZ camera 110 photographs each surveillance region to distinguish all moving objects. The control unit 140 stores data of all the distinguished objects in the storage unit 120 ).

The monitoring object selecting step S30 selects an object to be monitored for each monitoring area. In the monitoring object selecting step S30, the controller 140 selects an object to be monitored. At this time, the controller 140 may select different objects for respective areas. For example, as shown in FIG. 4C, the controller 140 selects a black passenger car as a target object in region b, selects a blue bus as a target object in region h, and selects a person who rides a motorcycle in region j You can select it as the target object.

The first learning step S40 is a step of classifying and learning the monitored object in each area. In the first learning step S40, the classifying unit 130 classifies the information of the monitored object among the data stored in the storage unit 120 using the classifier F. FIG. The classifying unit 130 classifies data selected as an object to be monitored among the data of the respective regions stored in the controller 120 into the selection data A and classifies the data that are not selected as objects to be monitored as non-selection data B , The selection data A and the non-selection data B are input to the classifier F so as to produce the output as shown in the following equation (1).

[Equation 1]

F (A) = 1, F (B) = - 1

That is, the classifier F performs learning to output 1 when selection data A, which is an object to be monitored, is input, and outputs -1 when non-selection data B that is not an object to be monitored is input.

The second extracting step (S50) is a step of re-extracting all objects moving in the monitoring area. In the second extraction step S50, the PTZ camera 110 photographs each surveillance area to distinguish all moving objects. The control unit 140 stores the data C of all the distinguished objects (120).

The second learning step S60 is a step of learning how to reclassify the monitored object among the stored data C in the second extracting step S50. In the second learning step S60, the classifying unit 130 uses the classifier F learned in the first learning step S40 to monitor the data C stored in the second extracting step S50 Learning to classify the information of the object is performed. That is, in the second learning step S60, the input data of the learned classifier is used as the input data C extracted in the second extraction step S50 (S61).

First, the classifier (F) reclassifies data using the following equation (2).

&Quot; (2) "

if F (C) > 0 then AA = AU C

if F (C) < 0 then BB = BI C

If the output of the classifier F is larger than 0 (S62) when the data C stored in the second extraction step S50 is input to the classifier F, And classifies it as selection data AA (S63). Therefore, the data A classified in the first learning step S40 and the data C classified in the second learning step S60 are stored in the selection data AA (S63).

When the data C stored in the second extraction step S50 is input to the classifier F and the output of the classifier F is smaller than 0 in step S62, It is recognized that the object is not an object to be monitored and classified into non-selection data BB (S64). Therefore, the data B classified in the first learning step S40 and the data C classified in the second learning step S60 are stored in the non-selection data BB in step S64.

Then, the classifier 130 inputs the selection data AA and the non-selection data BB to the classifier F and re-learns the output as shown in the following equation (3).

F (AA) = 1, F (BB) = - 1

That is, the classifier F outputs 1 when the selection data AA, which is an object to be monitored, is input, and re-learns to output -1 when non-selection data BB that is not an object to be monitored is input.

The monitoring start determination step S70 is a step of determining whether or not monitoring of the monitored object can be started. In the monitoring start determination step S70, the controller 140 compares the amount of data classified into the selected data AA with the reference value in the second learning step S60 (S71). If the amount of data classified as the selected data AA is larger than the reference value, the controller 140 determines that the data need not be collected any more and starts monitoring the object to be monitored by the PTZ camera 110 (S72). That is, the controller 140 determines that the PTZ camera 110 has learned enough to recognize the target object, and instructs the PTZ camera 110 to start monitoring the target object.

In step S70, the controller 140 compares the amount of data classified into the data AA selected in the second learning step S40 with a reference value in step S71. If the amount of data classified into the selected data AA is smaller than the reference value, the controller 140 determines that the amount of collected data is small and returns to the second extracting step S50 to perform a second learning step S60 ). That is, the control unit 140 determines that the PTZ camera 110 has not learned enough to recognize the object, and instructs the PTZ camera 110 to repeat the learning to classify the object.

As described above, the surveillance method of the video surveillance apparatus according to an embodiment of the present invention can automatically learn objects to be watched by performing the class teacher learning through the classifier.

In addition, the monitoring method of the video surveillance apparatus according to an embodiment of the present invention can automatically monitor multiple objects for each surveillance area by using a PTZ camera.

As described above, the present invention is not limited to the above-described embodiments, but can be applied to a video surveillance apparatus according to the present invention It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention.

110: PTZ camera
120:
130:
140:

Claims (9)

A surveillance area setting step of setting a surveillance area to be photographed by a PTZ (pan / tilt zoom) camera;
A first extraction step in which the PTZ camera captures a surveillance area and extracts data of all moving objects;
A control object selection step of selecting an object to be monitored to be monitored in the surveillance area;
A first learning step of classifying data of an object of observation among data of the object extracted in the first extraction step and learning the classifier through the classified data;
A second extracting step of re-extracting data of all moving objects by re-capturing the surveillance area by the PTZ camera; And
A second learning step of inputting the data of the object extracted in the second extraction step to the classifier learned in the first learning step to reclassify data of the object to be monitored and re-learning the classifier through the reclassified data Lt; / RTI &gt;
And monitoring the monitoring target object if the data amount of the monitoring target object is larger than a reference value. The method according to claim 1,
Wherein the first learning step classifies the data of the monitoring object into the selection data and classifies the data of the object other than the monitoring object into the non-selection data. 3. The method of claim 2,
Wherein when the selection data is input to the classifier, the first learning step outputs 1, and when the non-selection data is input to the classifier, the classifier learns to output -1. . The method according to claim 1,
Wherein in the second learning step, when the data of the object extracted in the second extraction step is input to the classifier, if the output of the classifier is larger than 0, the classification is reclassified as selection data. 5. The method of claim 4,
Wherein the second learning step re-learns the classifier to output 1 when the selection data is input to the classifier. The method according to claim 1,
And if the output of the classifier is less than 0 when the data of the object extracted in the second extracting step is input to the classifier in the second learning step, Way. The method according to claim 6,
And in the second learning step, the classifier is re-learned so as to output -1 when the non-selection data is input to the classifier. The method according to claim 1,
And if the data amount of the monitored object is smaller than the reference value, returning to the second extracting step and repeating the second learning step. The method according to claim 1,
Wherein the surveillance area setting step sets the surveillance area to a plurality of surveillance areas, and in the surveillance target object selection step, different surveillance targets are selected for each surveillance area, and the plurality of surveillance target objects are simultaneously monitored. Monitoring method.

Images