Area and rule automatic configuration monitoring method and monocular security camera
Through depth estimation algorithm and deep learning algorithm, identify and track potential dangerous areas, optimize monitoring rules, and solve the problem that monocular security cameras cannot distinguish between depth information and monitoring areas, improve identification accuracy and alarm accuracy, and reduce missed reports and false alarms.
Patent Information
- Application Number
- CN202411334974.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2025-01-03
AI Technical Summary
The monocular security camera cannot distinguish in-depth information, resulting in frequent irrelevant alarms. Moreover, due to the large monitoring area, it is prone to missed or false alarms, and the monitoring area cannot be flexibly controlled, affecting the recognition accuracy.
Depth estimation algorithm and deep learning algorithm are used to obtain the depth information of the monitoring screen, identify and track potential dangerous areas, optimize monitoring rules through machine self-learning, eliminate non-essential monitoring areas, and reduce alarm frequency.
Flexible configuration of monitoring areas within the depth range, improve identification accuracy, reduce the rate of missed and false alarms, dynamically optimize monitoring rules, and avoid unnecessary alarms.
Smart Images
Figure CN119251762A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of security camera design, and specifically relates to a method for automatically configuring monitoring of an area and its rules, and a monocular security camera. Background Art
[0002] Current monocular security cameras can only obtain two-dimensional monitoring images, cannot distinguish depth information, and cannot flexibly adjust the monitoring area within the depth range, resulting in frequent irrelevant alarms. For example, in order to monitor intrusion at one's own door, alarms frequently occur when there is intrusion at the neighbor's door.
[0003] Current monocular security cameras usually cover a larger monitoring area, resulting in a larger non-essential monitoring area in the monitoring screen. When identifying specific targets and judging events, they are easily interfered by a large number of irrelevant factors, which reduces the recognition accuracy. In addition, data processing has high requirements for equipment performance, and is prone to missed or false alarms. For example, the dangerous areas monitored at the door of your home should be the stairwells and elevator entrances, not the wall side.
[0004] Current monocular security cameras send alarms based on preset rules, but the preset rules are mostly simple and unreasonable, and are prone to missed or false alarms. For example, it is stipulated that an alarm should be sent when someone appears at the door of your home, at the staircase, or at the elevator entrance, but it is not desired to send an alarm when someone at your home or a common type of non-dangerous person appears. It is also not desired to send an alarm in common scenarios where there is no danger in appearing at the staircase or elevator entrance for only a short stay.
[0005] This application is proposed in view of the above-mentioned technical defects. Summary of the invention
[0006] The purpose of this application is to provide a method for automatically configuring monitoring of an area and its rules and a monocular security camera to overcome or alleviate at least one of the known technical defects.
[0007] The technical solution of this application is:
[0008] On the one hand, a method for automatically configuring monitoring of an area and its rules is provided, including: a monitoring screen acquisition step, a monitoring screen preprocessing step, a monitoring screen depth processing step, a potential danger area identification step, a potential danger area tracking step, a monitoring screen storage step, a monitoring rule optimization step, and an alarm step, wherein:
[0009] Monitoring screen acquisition steps: Acquire monitoring screen;
[0010] Monitoring screen preprocessing steps: preprocess the monitoring screen;
[0011] Depth processing steps of monitoring images: Use depth estimation algorithm to estimate the depth of monitoring images and obtain the depth information of pixels;
[0012] Potentially dangerous area identification step: Use deep learning algorithms to understand the scene in the monitoring screen and identify potential dangerous areas in the monitoring screen;
[0013] Steps for tracking potential danger areas: Track potential danger areas;
[0014] Monitoring screen storage steps: storing monitoring screen;
[0015] Monitoring rule optimization steps: extract scene features from historical monitoring images, conduct self-learning using machine self-learning methods, deeply understand the dynamic change characteristics of the scene, exclude common non-dangerous intrusion scenarios, and optimize monitoring rules;
[0016] Alarm steps: When the monitoring screen matches the monitoring rules, an alarm is issued.
[0017] Optionally, in the above-mentioned method for automatically configuring monitoring of regions and their rules, in the monitoring picture preprocessing step, the monitoring picture is preprocessed, including denoising and smoothing.
[0018] Optionally, in the above-mentioned method for automatically configuring and monitoring areas and their rules, in the step of tracking potential dangerous areas, if there are multiple potential dangerous areas at the same time, they are tracked simultaneously.
[0019] Optionally, in the above-mentioned method for automatically configuring monitoring of areas and their rules, in the monitoring rule optimization step, while optimizing the monitoring rules, non-dangerous areas are defined;
[0020] In the step of identifying the potential danger area, when identifying the potential danger area in the monitoring image, the non-dangerous area is not identified as the potential danger area.
[0021] Optionally, in the above-mentioned method for automatically configuring monitoring of areas and their rules, in the alarm step, an alarm is issued, which is a sound alarm or a light alarm.
[0022] On the other hand, a monocular security camera with automatic configuration of regions and their rules is provided, including: a monitoring picture acquisition module, a monitoring picture preprocessing module, a monitoring picture depth processing module, a potential danger area identification module, a potential danger area tracking module, a monitoring picture storage module, a monitoring rule optimization module, and an alarm module, wherein:
[0023] Monitoring screen acquisition module: collect monitoring screen;
[0024] Monitoring screen preprocessing module: preprocess the monitoring screen;
[0025] Monitoring screen depth processing module: Use the depth estimation algorithm to estimate the depth of the monitoring screen and obtain the depth information of the pixels;
[0026] Potentially dangerous area identification module: understands the scene in the monitoring screen through deep learning algorithms and identifies potential dangerous areas in the monitoring screen;
[0027] Potentially dangerous area tracking module: Tracking potential dangerous areas;
[0028] Monitoring screen storage module: store monitoring screen;
[0029] Monitoring rule optimization module: extracts scene features from historical monitoring images, performs self-learning using machine self-learning methods, deeply understands the dynamic change characteristics of scenes, excludes common non-dangerous intrusion scenarios, and optimizes monitoring rules;
[0030] Alarm module: issues an alarm when the monitoring screen matches the monitoring rules.
[0031] Optionally, the above-mentioned areas and their rules are automatically configured in a monocular security camera, and in a monitoring image preprocessing module, the monitoring image is preprocessed, including denoising and smoothing.
[0032] Optionally, the above-mentioned areas and their rules are automatically configured in a monocular security camera, and in a potential danger area tracking module, if there are multiple potential danger areas at the same time, they are tracked simultaneously.
[0033] Optionally, in the above-mentioned area and its rules automatic configuration monocular security camera, in the monitoring rule optimization module, while optimizing the monitoring rules, non-dangerous areas are defined;
[0034] In the potential danger area identification module, when identifying the potential danger area in the monitoring picture, the non-dangerous area is not identified as the potential danger area.
[0035] Optionally, the above-mentioned areas and their rules are automatically configured in a monocular security camera, and an alarm is issued in an alarm module, which is a sound alarm or a light alarm.
[0036] This application has at least the following beneficial technical effects:
[0037] A method for automatically configuring monitoring of an area and its rules and a monocular security camera are provided. A depth estimation algorithm is designed to perform depth estimation on the monitoring image and obtain depth information of pixels, so that the monitoring area can be flexibly configured within the depth range, and irrelevant monitoring areas can be excluded in depth. The scene in the monitoring image can be understood through a deep learning algorithm, and potential dangerous areas in the image can be identified, and then the potential dangerous areas can be dynamically tracked, so that non-essential monitoring areas can be excluded as much as possible in terms of scope. In addition, the design extracts scene features of historical monitoring images, performs self-learning with a machine self-learning method, deeply understands the dynamic change characteristics of the scene, excludes common non-dangerous intrusion scenes, and dynamically optimizes monitoring rules, thereby avoiding missed alarms or false alarms. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 It is a schematic diagram of a method for automatically configuring and monitoring areas and their rules provided in an embodiment of the present application;
[0039] Figure 2 It is a schematic diagram of the area and its rule-based automatic configuration of a monocular security camera provided in an embodiment of the present application.
[0040] In order to better illustrate the present embodiment, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product. In addition, the drawings are only used for illustrative purposes and should not be construed as limitations on the present application. DETAILED DESCRIPTION
[0041] In order to make the technical solution and advantages of the present application clearer, the technical solution of the present application will be described in further detail in detail and in detail with reference to the accompanying drawings. It can be understood that the specific embodiments described here are only partial embodiments of the present application, which are only used to explain the present application, not to limit the present application. It should be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, and other related parts can refer to the general design.
[0042] In addition, unless otherwise defined, the technical terms or scientific terms used in the description of this application shall have the usual meanings understood by those of ordinary skill in the art to which this application belongs. The words used in the description of this application to indicate orientation are only used to indicate relative directions or positional relationships. When the absolute position of the described object changes, its relative positional relationship may also change accordingly. The word "include" used in the description of this application indicates that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, but does not exclude other elements or objects.
[0043] In addition, it should be noted that, unless otherwise clearly specified and limited, the words "installation", "connection" and similar terms used in the description of this application should be understood in a broad sense. For example, the connection can be a fixed connection or a detachable connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. Technical personnel in the field can understand its specific meaning in this application according to the specific circumstances.
[0044] A method for automatically configuring monitoring of regions and their rules includes a monitoring screen acquisition step, a monitoring screen preprocessing step, a monitoring screen depth processing step, a potential danger zone identification step, a potential danger zone tracking step, a monitoring screen storage step, a monitoring rule optimization step, and an alarm step. Figure 1 shown.
[0045] Monitoring screen collection steps: Collect monitoring screen.
[0046] Monitoring screen preprocessing steps: Preprocess the monitoring screen, including denoising, smoothing, etc.
[0047] Depth processing steps of monitoring images: Use the depth estimation algorithm to estimate the depth of the monitoring image, obtain the depth information of the pixels, and output the 3D detection frame of the object. This allows the monitoring area to be flexibly configured within the depth range and exclude areas that are irrelevant to monitoring in terms of depth.
[0048] The depth estimation algorithm can use Depth Anything or MiDaS.
[0049] Steps for identifying potential dangerous areas: Use deep learning algorithms to understand the scenes in the surveillance images and identify potential dangerous areas in the surveillance images. For example, focus on areas where people may be active, such as stairways and elevator entrances, and exclude areas such as walls and windows.
[0050] Image segmentation models and scene understanding models can be used to identify potential dangerous areas in monitoring images.
[0051] Steps for tracking potential danger areas: Track potential danger areas and collect monitoring images to exclude unnecessary monitoring areas as much as possible.
[0052] If there are multiple potential danger areas at the same time, they can be tracked simultaneously.
[0053] Monitoring screen storage steps: store monitoring screen.
[0054] Monitoring rule optimization steps: extract scene features from historical monitoring images, conduct self-learning using machine self-learning methods, deeply understand the dynamic change characteristics of the scene, exclude common non-dangerous intrusion scenarios, optimize monitoring rules, and define non-dangerous areas.
[0055] At the same time, considering that actual dangerous intrusion is a low-probability event, in the potential dangerous area identification step, when identifying the potential dangerous area in the monitoring screen, the non-dangerous area is not identified as a potential dangerous area. This can reduce the frequency of tracking the potential dangerous area, narrow the tracking scope, and improve the identification accuracy.
[0056] The machine self-learning method can be implemented by combining memory with agent scheduling.
[0057] Alarm steps: When the monitoring screen matches the monitoring rules, an alarm is issued, which can be a sound alarm or a light alarm. When extracting events in the monitoring screen, image segmentation models and scene understanding models can be used to ensure the recognition accuracy of specific targets.
[0058] A monocular security camera with automatic configuration of regions and their rules includes a monitoring picture acquisition module, a monitoring picture preprocessing module, a monitoring picture depth processing module, a potential danger area identification module, a potential danger area tracking module, a monitoring picture storage module, a monitoring rule optimization module, and an alarm module. Figure 2 shown.
[0059] Monitoring screen acquisition module: collect monitoring screen.
[0060] Monitoring screen preprocessing module: preprocess the monitoring screen, including denoising, smoothing, etc.
[0061] Monitoring screen depth processing module: Use the depth estimation algorithm to estimate the depth of the monitoring screen, obtain the depth information of the pixels, and output the 3D detection frame of the object. This allows the monitoring area to be flexibly configured within the depth range and exclude areas that are irrelevant to monitoring in terms of depth.
[0062] The depth estimation algorithm may adopt algorithms such as Depth Anything and MiDaS.
[0063] Potentially dangerous area identification module: Use deep learning algorithms to understand the scenes in the monitoring screen and identify potential dangerous areas in the monitoring screen. For example, focus on areas where people may be active, such as stairways and elevator entrances in the monitoring screen, and exclude areas such as walls and windows.
[0064] Image segmentation models and scene understanding models can be used to identify potential dangerous areas in monitoring images.
[0065] Potentially dangerous area tracking module: Tracks potential dangerous areas and collects monitoring images to exclude unnecessary monitoring areas as much as possible.
[0066] If there are multiple potential danger areas at the same time, they can be tracked simultaneously.
[0067] Monitoring screen storage module: stores monitoring screens.
[0068] Monitoring rule optimization module: extracts scene features from historical monitoring images, performs self-learning using machine self-learning methods, deeply understands the dynamic change characteristics of the scene, excludes common non-dangerous intrusion scenarios, optimizes monitoring rules, and defines non-dangerous areas.
[0069] At the same time, considering that actual dangerous intrusion is a low-probability event, in the potential dangerous area identification module, when identifying the potential dangerous area in the monitoring screen, the non-dangerous area is not identified as a potential dangerous area. This can reduce the working frequency of the potential dangerous area tracking module to control the tracking of the potential dangerous area, extend the overall life of the system, narrow the tracking range, and improve the identification accuracy.
[0070] The machine self-learning method can be implemented by combining memory with agent scheduling.
[0071] Alarm module: When the monitoring screen matches the monitoring rules, an alarm is issued, which can be a sound alarm or a light alarm. When extracting events in the monitoring screen, image segmentation models and scene understanding models can be used to ensure the recognition accuracy of specific targets.
[0072] The above-mentioned embodiment discloses a monocular security camera that automatically configures the area and its rules, which corresponds to the above-mentioned embodiment discloses a monitoring method that automatically configures the area and its rules, wherein each module can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, this application generally describes them according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Technical personnel in the field can choose to adopt different methods to implement the described functions for each specific application and its actual constraints, but such implementation should not be considered to exceed the scope of this application.
[0073] The method for automatically configuring monitoring areas and their rules and the monocular security camera disclosed in the above-mentioned embodiments are designed with a depth estimation algorithm to perform depth estimation on the monitoring screen and obtain depth information of pixels, so that the monitoring area can be flexibly configured within the depth range, and irrelevant monitoring areas can be excluded in depth. The scene in the monitoring screen can be understood through a deep learning algorithm, and the potential dangerous areas in the screen can be identified, and then the potential dangerous areas can be dynamically tracked, so that non-essential monitoring areas can be excluded as much as possible in terms of scope. In addition, the design extracts scene features of historical monitoring screens, performs self-learning with a machine self-learning method, deeply understands the dynamic change characteristics of the scene, excludes common non-dangerous intrusion scenes, and dynamically optimizes the monitoring rules, thereby avoiding missed alarms or false alarms.
[0074] So far, the technical solution of the present application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. Those skilled in the art should understand that the scope of protection of the present application is obviously not limited to these specific embodiments. Without departing from the principles of the present application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the scope of protection of the present application.
Claims
1. A method for automatically configuring and monitoring a region and its rules, characterized in that: include: Monitoring screen acquisition step, monitoring screen preprocessing step, monitoring screen depth processing step, potential danger area identification step, potential danger area tracking step, monitoring screen storage step, monitoring rule optimization step, alarm step, among which, Monitoring screen acquisition steps: Acquire monitoring screen; Monitoring screen preprocessing steps: preprocess the monitoring screen; Depth processing steps of monitoring images: Use depth estimation algorithm to estimate the depth of monitoring images and obtain the depth information of pixels; Potentially dangerous area identification step: Use deep learning algorithms to understand the scene in the monitoring screen and identify potential dangerous areas in the monitoring screen; Steps for tracking potential danger areas: Track potential danger areas; Monitoring screen storage steps: storing monitoring screen; Monitoring rule optimization steps: extract scene features from historical monitoring images, conduct self-learning using machine self-learning methods, deeply understand the dynamic change characteristics of the scene, exclude common non-dangerous intrusion scenarios, and optimize monitoring rules; Alarm steps: When the monitoring screen matches the monitoring rules, an alarm is issued.
2. The method for automatically configuring and monitoring areas and their rules according to claim 1, characterized in that: In the monitoring picture preprocessing step, the monitoring picture is preprocessed, including denoising and smoothing.
3. The area and rule automatic configuration monitoring method according to claim 1 is characterized in that: In the potential danger area tracking step, if there are multiple potential danger areas at the same time, they are tracked simultaneously.
4. The method for automatically configuring and monitoring areas and their rules according to claim 1, characterized in that: In the monitoring rule optimization step, while optimizing the monitoring rules, non-dangerous areas are defined; In the step of identifying the potential danger area, when identifying the potential danger area in the monitoring image, the non-dangerous area is not identified as the potential danger area.
5. The area and rule automatic configuration monitoring method according to claim 1 is characterized in that: In the alarm step, an alarm is issued, which is an audible alarm or a light alarm.
6. A monocular security camera with automatic configuration of regions and their rules, characterized in that: include: Monitoring screen acquisition module, monitoring screen preprocessing module, monitoring screen depth processing module, potential danger area identification module, potential danger area tracking module, monitoring screen storage module, monitoring rule optimization module, alarm module, among which, Monitoring screen acquisition module: collect monitoring screen; Monitoring screen preprocessing module: preprocess the monitoring screen; Monitoring screen depth processing module: Use the depth estimation algorithm to estimate the depth of the monitoring screen and obtain the depth information of the pixels; Potentially dangerous area identification module: understands the scene in the monitoring screen through deep learning algorithms and identifies potential dangerous areas in the monitoring screen; Potentially dangerous area tracking module: Tracking potential dangerous areas; Monitoring screen storage module: store monitoring screen; Monitoring rule optimization module: extracts scene features from historical monitoring images, performs self-learning using machine self-learning methods, deeply understands the dynamic change characteristics of scenes, excludes common non-dangerous intrusion scenarios, and optimizes monitoring rules; Alarm module: issues an alarm when the monitoring screen matches the monitoring rules.
7. The area and rules according to claim 6 are used to automatically configure a monocular security camera, characterized in that: In the monitoring image preprocessing module, the monitoring image is preprocessed, including denoising and smoothing.
8. The area and rule-based automatic configuration of a monocular security camera according to claim 6, characterized in that: In the potential danger area tracking module, if there are multiple potential danger areas at the same time, they will be tracked simultaneously.
9. The area and rule-based automatic configuration of a monocular security camera according to claim 6, characterized in that: In the monitoring rule optimization module, non-dangerous areas are defined while optimizing monitoring rules; In the potential danger area identification module, when identifying the potential danger area in the monitoring screen, the non-dangerous area is not identified as the potential danger area.
10. The area and rule-based automatic configuration of a monocular security camera according to claim 6, characterized in that: In the alarm module, an alarm is issued, which is a sound alarm or a light alarm.