Gimbal camera hidden trouble tracking method

Abstract

The application discloses a hidden danger tracking method of a pan-tilt camera and belongs to the fields of image processing and computer vision. In the scene where a plurality of hidden danger targets are identified in a video picture, the application divides a monitoring picture into a plurality of monitoring areas, sets an area weight of each monitoring area, and sets a type weight of each hidden danger type. By comparing the hidden danger targets with each monitoring area, the position weight of each hidden danger target is calculated, the tracking weight of each hidden danger target is calculated by combining the type weight of each hidden danger target, and the final tracking hidden danger target is determined according to the maximum value of different tracking weights. The application is close to the process of manually selecting a hidden danger target to be tracked, and realizes the automatic and accurate selection of the hidden danger target to be tracked.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and computer vision, and particularly to a gimbal camera hidden danger tracking method. BACKGROUND

[0002] The gimbal camera includes a gimbal and a camera arranged on the gimbal, and the gimbal can rotate in multiple directions to change the monitoring area of the camera. The current gimbal camera can be used in various scenes such as daily hidden danger monitoring, construction site hidden danger monitoring, and power transmission line hidden danger monitoring. The gimbal camera of the prior art can identify hidden dangers in real time, and track the hidden danger target for a period of time after identifying the hidden danger target. However, the prior art has great randomness in automatically selecting a tracking target in a scene where multiple hidden danger targets exist, and the selected hidden danger target is often not the one that is finally desired to track, so manual selection of the tracking target is required. SUMMARY

[0003] The present application provides a gimbal camera hidden danger tracking method, which is close to the process of manually selecting a hidden danger target to be tracked in a scene where multiple hidden danger targets are identified in a video image, and realizes automatic and accurate selection of the hidden danger target to be tracked.

[0004] The present application provides the following technical solutions:

[0005] A gimbal camera hidden danger tracking method, the method comprising:

[0006] S10: dividing a monitoring image of the camera into multiple monitoring areas in advance, and assigning each monitoring area with a respective area weight according to the attention degree of each monitoring area to hidden dangers;

[0007] S20: assigning each hidden danger type with a respective type weight according to the attention degree of each hidden danger type and the severity of each hidden danger type;

[0008] S30: controlling the camera at an initial position by the gimbal, continuously shooting a video by the camera in real time, and performing hidden danger identification in real time, identifying hidden danger targets and determining the hidden danger type of each hidden danger target;

[0009] S40: when hidden danger tracking is required, comparing the position area of each hidden danger target with each monitoring area of the monitoring image, finding all monitoring areas that overlap with each hidden danger target, and taking the maximum value of the area weights of all monitoring areas that overlap with each hidden danger target as the position weight of each hidden danger target;

[0010] S50: calculating the tracking weight of each hidden danger target according to the position weight of each hidden danger target and the type weight of the hidden danger type of each hidden danger target;

[0011] S60: taking the hidden danger target with the largest tracking weight as a to-be-tracked target, controlling the camera by the pan-tilt to track the to-be-tracked target, and returning the camera to an initial position after tracking.

[0012] Further, if there are multiple hidden danger targets with the largest tracking weight, the multiple hidden danger targets with the largest tracking weight are taken as first-class candidate targets, and a hidden danger target with the largest position weight in the first-class candidate targets is selected as the to-be-tracked target.

[0013] If there are multiple hidden danger targets with the largest position weight in the first-class candidate targets, the multiple hidden danger targets with the largest position weight are taken as second-class candidate targets, and a hidden danger target with the largest type weight in the second-class candidate targets is selected as the to-be-tracked target.

[0014] If there are multiple hidden danger targets with the largest type weight in the second-class candidate targets, the multiple hidden danger targets with the largest type weight are taken as third-class candidate targets, and a hidden danger target with the largest hidden danger area in the third-class candidate targets is selected as the to-be-tracked target.

[0015] Further, when tracking the to-be-tracked target, the to-be-tracked target is enlarged to a set size, and the pan-tilt is rotated, so that the to-be-tracked target is always displayed on a center position of a monitoring picture.

[0016] Further, in the process of rotating the pan-tilt, a hidden danger target with the same hidden danger type as the to-be-tracked target and closest to a center point of the monitoring picture is selected to be tracked.

[0017] Further, in the process of rotating the pan-tilt, if there is no hidden danger target with the same hidden danger type as the to-be-tracked target, a hidden danger target closest to the center point of the monitoring picture is selected to be tracked.

[0018] Further, the tracking weight of each hidden danger target is calculated by the following formula:

[0019] Tracking weight = position weight + type weight.

[0020] The present application has the following beneficial effects:

[0021] In the scene of identifying multiple hidden danger targets in a video picture, the monitoring picture is divided into multiple monitoring areas, and the area weight of each monitoring area is set, and the type weight of each hidden danger type is set. By comparing the hidden danger target with each monitoring area, the position weight of each hidden danger target is calculated, and the tracking weight of each hidden danger target is calculated by the type weight of each hidden danger target. According to the maximum value of different tracking weights, the final tracked hidden danger target is determined. The process of manually selecting the hidden danger target to be tracked is close to reality, and the automatic and accurate selection of the hidden danger target to be tracked is realized. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A flowchart of the gimbal camera hidden danger tracking method of the application;

[0023] Figure 2 A schematic diagram of the multiple monitoring areas divided in the monitoring picture and the area weight thereof;

[0024] Figure 3 A schematic diagram of the comparison of the hidden danger target with each monitoring area. DETAILED DESCRIPTION

[0025] In order to make the technical problems, technical solutions and advantages of the application more clear, the following will be described in detail with reference to the drawings and specific embodiments.

[0026] The embodiment of the application provides a gimbal camera hidden danger tracking method, as shown in the figure, the method comprises the following steps: Figure 1

[0027] S10: The monitoring picture of the camera is divided into multiple monitoring areas in advance, and each monitoring area is assigned with a respective area weight according to the attention degree of the hidden danger of each monitoring area.

[0028] For example, as shown in the figure, the monitoring picture is divided into nine monitoring areas, and each monitoring area is provided with an area weight value. The greater the area weight, the higher the attention degree of the monitoring area. Figure 2

[0029] S20: Each hidden danger type is assigned with a respective type weight according to the attention degree of each hidden danger type and the severity of each hidden danger type.

[0030] The type of hidden danger can be various, such as construction machinery, fireworks, etc. The application assigns different type weight values to each different hidden danger type according to the attention and severity. The greater the type weight, the higher the attention degree of the hidden danger.

[0031] ​​S30: The PTZ controls the camera to be at the initial position. The camera continuously shoots video in real time and performs hidden danger identification in real time, identifies hidden danger targets and determines the hidden danger type of each hidden danger target.

[0032] In the present invention, the camera is set on a pan-tilt head, and the rotation of the camera is controlled by rotating the pan-tilt head up, down, left, and right. In this step, the pan-tilt head can be rotated to select a scene at an angle as the initial position for tracking (the initial position can be determined by setting the home position or preset position). The camera continuously shoots video in real time and starts hidden danger analysis and identification of the real-time video image, identifies hidden danger targets and determines the hidden danger type of each hidden danger target. The specific method of hidden danger identification adopts the method of the existing technology, such as machine learning and other methods, which will not be described in detail in the present invention.

[0033] S40: When hidden danger tracking is required, the location area of ​​each hidden danger target is overlapped with each monitoring area of ​​the monitoring screen to find all monitoring areas that overlap with each hidden danger target, and the maximum value of the area weights of all monitoring areas that overlap with each hidden danger target is used as the location weight of each hidden danger target.

[0034] In this step, when hidden danger tracking is needed, you can add a hidden danger tracking task and set a tracking time. The tracking task is considered complete when the time expires, for example, each tracking task can be tracked for 10 seconds. When performing hidden danger tracking tasks, the location of the hidden danger target in the monitoring screen is compared with the overlapping monitoring areas. The monitoring area with the highest area weight among the overlapping monitoring areas is selected and its area weight is used as the location weight of the hidden danger target.

[0035] S50: Calculate the tracking weight of each hidden danger target according to the position weight of each hidden danger target and the type weight of the hidden danger type of each hidden danger target.

[0036] Exemplarily, the tracking weight of each hidden danger target can be calculated by the following formula: tracking weight = location weight + type weight.

[0037] S60: The hidden danger target with the largest tracking weight is taken as the target to be tracked, and the pan-tilt control camera tracks the target to be tracked. After the tracking is completed, the pan-tilt control camera returns to the initial position.

[0038] like Figure 3 As shown in the figure, assuming that the dotted box represents the analyzed hidden danger targets, hidden danger targets 1, 2, and 3 are of the same hidden danger type, such as excavator, and their type hidden danger weights are all 6, then:

[0039] The tracking weight of hidden danger target 1 = 6 + 6 = 12;

[0040] Tracking weight of hidden danger target 2 = 9 + 6 = 15

[0041] Tracking weight of hidden danger target 3 = 5 + 6 = 11

[0042] Finally, hidden danger target 2 is selected as the target to be tracked. After the tracking task is completed (such as reaching the set tracking time), the initial position is returned, and the operations of S30-S60 are repeated to select the next target to be tracked for continuous tracking.

[0043] In the scenario of identifying multiple hidden danger targets in a video screen, the application divides the monitoring screen into multiple monitoring areas and sets the area weight of each monitoring area, and sets the type weight of each hidden danger type. By comparing the overlap of hidden danger targets and each monitoring area, the position weight of each hidden danger target is calculated, and the tracking weight of each hidden danger target is calculated with the type weight of each hidden danger target. According to the maximum value of different tracking weights, the final tracking hidden danger target is determined. The process of manually selecting the hidden danger target to be tracked in the application is close to reality, and the automatic and accurate selection of the hidden danger target to be tracked is realized.

[0044] In the foregoing selection of the target to be tracked, if there are multiple hidden danger targets with the maximum tracking weight, the multiple hidden danger targets with the maximum tracking weight are taken as the first type of candidate target, and the hidden danger target with the maximum position weight in the first type of candidate target is selected as the target to be tracked.

[0045] If there are multiple hidden danger targets with the maximum position weight in the first type of candidate target, the multiple hidden danger targets with the maximum position weight are taken as the second type of candidate target, and the hidden danger target with the maximum type weight in the second type of candidate target is selected as the target to be tracked.

[0046] If there are multiple hidden danger targets with the maximum type weight in the second type of candidate target, the multiple hidden danger targets with the maximum type weight are taken as the third type of candidate target, and the hidden danger target with the maximum hidden danger area in the third type of candidate target is selected as the target to be tracked.

[0047] That is, when there are multiple hidden danger targets with the maximum tracking weight, the hidden danger target with the maximum position weight is preferentially selected for tracking, the hidden danger target with the maximum type weight is then selected for tracking, and finally the hidden danger target with the maximum hidden danger area is selected for tracking.

[0048] In the foregoing S60 of the application, when tracking the target to be tracked, the target to be tracked is enlarged to a set size (such as 1 / 8 of the monitoring screen), and the gimbal is rotated so that the target to be tracked is always displayed at the center position of the monitoring screen.

[0049] The prior art is susceptible to other hidden hazards during tracking, thereby losing the tracking target. To solve the problem, the present application selects a hidden target of the same type as the hidden target to be tracked and closest to the center of the monitoring picture during rotation of the holder, and rotates the holder to ensure that the hidden target being tracked is always at the center of the picture until the tracking task is completed.

[0050] Further, during rotation of the holder, if there is no hidden target of the same type as the hidden target to be tracked, a hidden target closest to the center of the monitoring picture is selected for tracking until the present tracking task is completed.

[0051] During hidden hazard tracking, the present application tracks a hidden target of the same type closest to the center, and if there is no hidden target of the same type, tracks another hidden target closest to the center, so that the initially tracked hidden target can be tracked to the maximum extent during one tracking process, and the tracking target is prevented from being lost due to interference from other hidden hazards.

[0052] The above is a preferred embodiment of the present application, and it should be noted that those of ordinary skill in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered within the scope of the present application.

Claims

1. A method for tracking hidden dangers of a PTZ camera, characterized in that: The method comprises: S10: Divide the camera's surveillance image into multiple surveillance areas in advance, and assign each surveillance area a respective area weight according to the degree of concern for hidden dangers in each surveillance area; S20: pre-assigning a type weight to each hidden danger type according to the degree of concern for each hidden danger type and the severity of each hidden danger type; S30: The PTZ controls the camera to stay at the initial position. The camera continuously captures video in real time and performs hidden danger identification in real time, identifying hidden danger targets and determining the hidden danger type of each hidden danger target. S40: When hidden danger tracking is required, the location area of ​​each hidden danger target is compared with each monitoring area of ​​the monitoring screen to find all monitoring areas that overlap with each hidden danger target. The maximum value of the area weights of all monitoring areas that overlap with each hidden danger target is used as the location weight of each hidden danger target. S50: Calculating the tracking weight of each hidden danger target according to the position weight of each hidden danger target and the type weight of the hidden danger type of each hidden danger target; S60: The hidden danger target with the largest tracking weight is taken as the target to be tracked, and the pan-tilt control camera tracks the target to be tracked. After the tracking is completed, the pan-tilt control camera returns to the initial position.

2. The method for tracking hidden dangers of a PTZ camera according to claim 1, wherein: If there are multiple hidden danger targets with the largest tracking weights, the multiple hidden danger targets with the largest tracking weights are taken as first-category candidate targets, and the hidden danger target with the largest position weight among the first-category candidate targets is selected as the target to be tracked; If there are multiple potential danger targets with the largest position weights among the first category of candidate targets, then the multiple potential danger targets with the largest position weights are selected as the second category of candidate targets, and the potential danger target with the largest type weight among the second category of candidate targets is selected as the target to be tracked; If there are multiple hidden danger targets with the largest type weights among the second category candidate targets, the multiple hidden danger targets with the largest type weights are taken as the third category candidate targets, and the hidden danger target with the largest hidden danger area among the third category candidate targets is selected as the target to be tracked.

3. The method for tracking hidden dangers of a PTZ camera according to claim 1, wherein: When tracking the target to be tracked, the target to be tracked is magnified to a set size, and the pan / tilt head is rotated so that the target to be tracked is always displayed at the center of the monitoring screen.

4. The method for tracking hidden dangers of a PTZ camera according to claim 3, characterized in that: During the rotation of the pan / tilt head, a hidden danger target having the same hidden danger type as the target to be tracked and closest to the center point of the monitoring screen is selected for tracking.

5. The method for tracking hidden dangers of a PTZ camera according to claim 4, characterized in that: During the rotation of the pan / tilt head, if there is no hidden danger target of the same hidden danger type as the target to be tracked, the hidden danger target closest to the center point of the monitoring screen is selected for tracking.

6. The method for tracking hidden dangers of a PTZ camera according to any one of claims 1 to 5, characterized in that: The tracking weight of each hidden danger target is calculated using the following formula: Tracking weight = position weight + type weight.

Images