Electronic safety warning device and method based on image recognition
Through electronic safety warning devices based on image recognition, combined with high-definition cameras and cloud algorithms, real-time and comprehensive safety monitoring and early warning of power marketing operation sites are achieved, which solves the limitations of traditional supervision methods and improves the effectiveness and portability of safety management.
Patent Information
- Application Number
- CN202510848928.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-24
- Publication Date
- 2025-09-09
AI Technical Summary
Traditional safety supervision methods are difficult to achieve effective all-round and full-process monitoring at power marketing operation sites, and lack real-time safety warning functions, resulting in high safety management risks.
An electronic safety warning device based on image recognition is designed. It combines a high-definition camera, an infrared imaging light, and a cloud-based risk point identification algorithm. It monitors and marks risk points in real time through a wireless communication module, and supports portable installation and remote data interaction.
It achieves real-time and comprehensive security monitoring, ensuring monitoring effects without blind spots, and provides real-time warnings through portable devices and remote cloud analysis to reduce security management risks.
Smart Images
Figure CN120614510A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power grid inspection equipment, and in particular relates to an electronic safety warning device and method based on image recognition. Background Art
[0002] Safety supervision is a crucial aspect of power marketing operations and other production environments. However, traditional safety supervision approaches face numerous challenges. Due to the large number and widespread distribution of work sites, as well as the short duration of operations, power supply company safety supervisors struggle to conduct timely and comprehensive safety oversight of all sites. This not only hinders effective and timely implementation of safety organizational and technical measures, but also leads to frequent violations and persistently high safety management risks.
[0003] Although there have been some attempts to solve this problem in the existing technology, such as through on-site supervision and inspection and the installation of surveillance cameras for remote monitoring, these methods all have obvious limitations. On-site supervision and inspection rely on manual labor, which is not only costly, but also difficult to ensure long-term continuous supervision. Especially when there are many work sites and they are widely distributed, supervisors are prone to supervision loopholes due to fatigue or negligence. Although surveillance cameras can achieve remote video monitoring, they are large in size, require fixed installation, have a limited monitoring range, and are easily affected by signal strength, making it impossible to achieve effective monitoring of the entire process in all directions. In addition, surveillance cameras lack work safety protection and early warning functions, and cannot provide real-time safety reminders and early warnings to on-site workers. Summary of the Invention
[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides an electronic safety warning device and method based on image recognition to solve the above-mentioned technical problems.
[0005] In a first aspect, the present invention provides an electronic safety warning device based on image recognition, comprising a housing and a magnetic back clip, wherein a display screen is provided on the front of the housing, a buckle is provided above the display screen, function key areas are provided on the left and right sides of the housing, a camera is installed on the back of the housing, infrared imaging lights are provided on the upper and lower sides of the camera, and a wireless communication module, an integrated relay module, a circuit board containing a controller, and a battery for powering the entire device are provided inside the housing; the buckle is connected to the magnetic back clip, and the magnetic back clip can be worn on the chest of a staff member or adsorbed on iron objects in the work area; The camera is connected to the input end of the controller, and the camera, infrared imaging light and display screen are all connected to the output end of the controller. The controller uploads the dynamic or static images taken by the camera to the cloud through the wireless communication module, identifies the risk points of the uploaded images through the risk point identification algorithm pre-stored in the cloud, and marks the identified risk points on the images. The images marked with risk points are transmitted to the display screen through the wireless communication module for display warnings.
[0006] A further improvement of the technical solution is that it also includes a locator installed in the shell, which is connected to the cloud via a wireless communication module.
[0007] A further improvement of the present technical solution is that the function key area is provided with a thermal imaging start key and an infrared imaging light start key, both of which are connected to the input end of the controller, the thermal imaging start key is used to control whether the camera is in the thermal imaging working mode, and the infrared imaging light start key is used to control whether the infrared imaging light is turned on.
[0008] A further improvement of this technical solution is that it also includes a Tap-c interface for charging the battery or for connecting an external mobile camera.
[0009] Further improvements to this technical solution include a Beidou communication start key, which is connected to the input end of the controller. The wireless communication module adopts the Beidou communication module, and the Beidou communication start key is used to control the opening and closing of the Beidou communication module.
[0010] In a second aspect, the present invention provides an electronic safety warning method based on image recognition, applicable to any of the electronic safety warning devices based on image recognition described above, comprising: After the device is turned on, the operator logs in by scanning the QR code on the security pass worn by the operator through the camera, and then selects the working mode after logging in; After selecting the working mode, the camera will take real-time photos of the work site and transmit the captured images to the cloud through the wireless communication module; The cloud identifies risk points in the received images based on pre-stored risk point identification algorithms; The cloud marks the identified risk points and transmits the images marked with risk points back to the display screen on the device through the wireless transmission module for displaying warnings.
[0011] A further improvement of this technical solution is that the pre-stored risk point identification algorithm adopts the SIFT algorithm. The SIFT algorithm compares the features of the received image with the risk point standard image pre-stored in the cloud database to confirm whether there are risk points in the received image.
[0012] A further improvement of this technical solution is that the method of identifying risk points on the received image using the SIFT algorithm specifically includes: Apply Gaussian function to blur and downsample the image captured by the camera, construct a Gaussian pyramid, form a scale space, and detect and confirm several key point candidates in the scale space; Perform Taylor series expansion on each key point candidate to locate its position and scale; Remove low-contrast points and edge response points from the key point candidates to determine the key points; For each key point, a 16x16 neighborhood is taken around it with each key point as the center, and it is divided into 4x4 sub-blocks to determine the neighborhood sampling area; Sampling is performed in the determined neighborhood sampling area, and the gradient amplitude and gradient direction of each sampling point are calculated; Generate a direction histogram in the neighborhood sampling area of the key point based on the calculated gradient amplitude and gradient direction, and determine the main direction of the key point; Calculate the gradient histogram in 8 directions for each sub-block in the neighborhood sampling area to form an 8-bin histogram. For each key point, obtain a 128-dimensional feature vector to form a key point descriptor; The extracted key point descriptors are compared with the features of the risk point standard images stored in the cloud database.
[0013] A further improvement of this technical solution is that the formula for calculating the gradient amplitude of the sampling point is: ; Among them, m(x, y) is the gradient amplitude of the sampling point, L(x+1, y) is the pixel value of the image at point (x+1, y), L(x-1, y) is the pixel value of the image at point (x-1, y), L(x, y+1) is the pixel value of the image at point (x, y+1), and L(x, y-1) is the pixel value of the image at point (x, y-1).
[0014] A further improvement of this technical solution is that the formula for calculating the gradient direction of the sampling point is: ; in, is the gradient direction of the sampling point.
[0015] The beneficial effects of the present invention are: Real-time, comprehensive safety monitoring: Using high-definition cameras and infrared imaging lights, the device captures real-time images of the worksite and provides night vision in dim conditions, ensuring comprehensive safety monitoring. Combined with pre-stored risk point recognition algorithms in the cloud, uploaded images are identified and labeled for risk points, providing real-time warnings and alerts of safety risks at the worksite.
[0016] Portable and Flexible Installation and Wear: The device is compact and lightweight, making it easy to carry and install. It can be worn on the chest or attached to iron objects in the work area via a magnetic clip, adapting to various working environments. The snap-on design allows the device to be connected to other fixing devices, such as regular clips, pins, and spring clips, further enhancing its flexibility.
[0017] Efficient data interaction and remote monitoring: The device exchanges data with carriers, the master control platform, and the cloud via wireless communication modules. It supports real-time positioning and manual or automatic network switching, ensuring stable and efficient data transmission. Furthermore, a relay interconnect button activates the centralized relay module, enhancing the device's signal reception capability and ensuring efficient data transmission even in poor signal conditions. A cloud database identifies and analyzes risk points in received images, transmitting the results back to the device's display screen for display warnings, enabling remote monitoring and real-time feedback.
[0018] In addition, the present invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0020] Figure 1 It is a structural diagram of the electronic safety warning device.
[0021] Figure 2 This is a schematic diagram of the structure of the electronic safety warning device after connecting the magnetic back clip.
[0022] Figure 3 This is a structural diagram of a wired mobile camera.
[0023] Figure 4 is a schematic flow chart of a method according to an embodiment of the present invention.
[0024] Figure 5 This is the main menu interface of the display screen.
[0025] Figure 6 This is the device login interface.
[0026] Figure 7 It is an electronic map interface.
[0027] Figure 8 It is the public network intercom interface.
[0028] Figure 9 This is the image shooting interface.
[0029] Figure 10 Set up the interface for electronic security fences.
[0030] Figure 11 Set up the interface for image capture.
[0031] Figure 12 It is the remote intelligent connection interface.
[0032] Figure 13 This is the inspection menu interface.
[0033] Figure 14 This is the image review interface.
[0034] Figure 15 It is the smart assistant interface.
[0035] Figure 16 This is the personal information interface.
[0036] Figure 17 This is the call recording interface.
[0037] Figure 18 This is the message center interface.
[0038] Figure 19 This is the local settings interface. DETAILED DESCRIPTION
[0039] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.
[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.
[0041] The key terms appearing in the present invention are explained below.
[0042] Image: It can be presented in two modes: static image and dynamic image. Static image is completed by taking photos, and dynamic image is completed by recording videos.
[0043] Electronic database: A database equipped with search, retrieval, data manipulation, and modeling capabilities for various safety risk points. This online system provides safety protection devices with various safety risk point identifiers, allowing them to identify existing safety risk points through captured or recorded images.
[0044] Power marketing: refers to the power companies' efforts to meet people's power consumption needs in a changing market environment, through a series of market-related business activities, to provide power products and corresponding services that meet consumer needs, thereby achieving the company's goals.
[0045] Power marketing on-site operations: refers to the work of conducting on-site surveys for business expansion and registration, installing meters and connecting electricity, power usage inspections, acceptance and acceptance of power transmission, data collection, operation and maintenance commissioning, etc. for electricity users. For power companies, only by ensuring the safety of customers' electrical equipment, the safety of the power grid, and personal safety are the primary conditions for their development. Therefore, safety management and control of power marketing on-site operations is particularly important.
[0046] On-site investigation: Check the scope of power outage required for on-site operations, retained live parts, location of grounding wires, adjacent lines, multiple power sources, self-contained power sources, underground pipeline facilities, conditions of the work site, environment and other dangerous points that affect the operation, and propose targeted safety measures and precautions.
[0047] Meter installation and power connection: This refers to the process whereby staff connect the energy meter to the customer's receiving device circuit. Once the meter installation is complete and confirmed to be correct, power can be inspected and delivered. The customer then turns on their own switch and begins using electricity.
[0048] Electricity inspection: Electricity inspection refers to the inspection, supervision, guidance and assistance carried out by power companies to ensure normal power supply and use order and public safety for users to use electricity safely, economically and reasonably.
[0049] Collection and operation maintenance: refers to the remote or on-site work performed by collection and operation maintenance personnel, including maintenance of the operating status of on-site collection devices and metering devices, replacement of SIM cards and communication modules, and upgrade of solid-state hardware versions, so as to ensure reliable operation of the collection devices, stable communication, and accurate data.
[0050] Collection debugging: refers to the failure of data collection. Through troubleshooting on the system side and the site side, the cause of the data failure is analyzed, and relevant maintenance measures are collected to ensure that the system can realize data collection.
[0051] like Figure 1 and Figure 2 As shown, the present invention provides an electronic safety warning device based on image recognition, including a shell and a magnetic back clip, a display screen is provided on the front of the shell, a buckle is provided above the display screen, function key areas are provided on the left and right sides of the shell, a camera is installed on the back of the shell, infrared imaging lights are provided on the upper and lower sides of the camera, and a wireless communication module, an integrated relay module, a circuit board containing a controller and a battery for powering the entire device are provided inside the shell; the buckle is connected to the magnetic back clip, and the magnetic back clip can be worn on the chest of the staff or adsorbed on iron objects in the work area; the camera is connected to the input end of the controller, and the camera, infrared imaging light and display screen are all connected to the output end of the controller, and the controller uploads the dynamic image or static image taken by the camera to the cloud through the wireless communication module, identifies the risk points of the uploaded image through the risk point identification algorithm pre-stored in the cloud, and marks the identified risk points on the image, and transmits the image marked with the risk points to the display screen through the wireless communication module for display warning.
[0052] Specifically, the device is 10cm long, 4.5cm wide, and 1cm thick, making it compact, lightweight, and easy to install and carry. It features a 2.4-inch high-definition display supporting 1080P high-definition image quality, and can capture real-time images of the work site. Risk points are then displayed in real-time on the images through database comparison.
[0053] Among them, the buckle can be connected to a magnetic back clip, an ordinary back clip, a pin, a spring clip, etc. The device can be worn on the chest of the operator through the magnetic back clip, or the magnetic back clip and the device can be combined and placed on surrounding iron objects for adsorption (such as metering screens, cabinets) and fixation. Combined with the on-site working environment, multi-angle and all-round retrieval can be achieved by recording high-definition video, shooting high-definition images, and automatic capture of first-person or third-person images. The magnetic back clip includes a back clip that is snapped onto the shell and a magnet set with the back clip. When in use, the staff snaps the shell onto the back clip through the buckle, places the magnet on the inside of the staff's work clothes, and places the back clip with the shell on the outside of the work clothes. After aligning the magnet and the back clip and combining them for adsorption, the equipment can be fixed.
[0054] In addition, the function key area is equipped with a power button, a selection button, a thermal imaging start button, an infrared imaging light start button, and a network switch button. The selection button, thermal imaging start button, infrared imaging light start button, and network switch button are all connected to the input terminal of the controller. The power button is used to turn on the device with one click; the selection button, including two upper and lower selection buttons, can select functions through the displayed menu; the thermal imaging start button is used to control whether the camera is in thermal imaging working mode, and the infrared imaging light start button is used to control whether the infrared imaging light is turned on. Specifically, the thermal imaging start button can turn on thermal imaging to further determine whether there are potential danger points such as bird nests, animals, and bare wires in the internal space of the operation; the infrared imaging light start button can control the infrared imaging light to turn on, providing night vision effects, facilitating the automatic capture of static images on the scene under dim conditions, and prompting potential danger points in the captured images; the network switch button can switch to the best wireless operator network according to the network conditions at the operation site to maintain efficient signal strength.
[0055] In addition, the device's shell is also equipped with a menu key, confirmation key, return key, SOS key, relay interconnection key, Beidou communication activation key, SIM card slot and Tap-c interface.
[0056] Menu key: One key can enter the function menu.
[0057] Confirmation key: A function key of the device, used for confirmation in the displayed menu.
[0058] Return key: A function key of the device, used to return in the displayed menu.
[0059] SOS button: used for emergency calls. When an emergency occurs at the work site, workers can press this button to call for help and record audio and video at the same time.
[0060] Relay interconnection key: used to turn on the centralized relay module in the device. When the signal strength at the work site cannot meet the requirements, the staff can press the relay interconnection key on site. The device can be networked and paired through the centralized relay module to enhance the device's ability to receive signals. Specifically, when the work site is located in a basement with no signal or indoors with poor signal, the relay interconnection key can be used to turn on the centralized relay module for signal relay transmission and signal amplification, with an effective transmission distance of up to 10 meters. The operator can set up a centralized relay module outdoors where the signal is better to receive strong outdoor signals. Along the path at the work site, a centralized relay module is placed every 10 meters or so for signal networking and transmission until the signal is received by the last centralized relay module.
[0061] Beidou communication start button: The Beidou communication start button is connected to the input end of the controller. The wireless communication module adopts the Beidou communication module. The Beidou communication start button is used to control the opening and closing of the Beidou communication module. It is suitable for open work sites or outdoor areas with signal drift.
[0062] SIM card slot: can be inserted with the wireless network device operator data card (adaptive to China Mobile, China Unicom, China Telecom signals, dual SIM dual standby), automatically match the communication parameters, and implement remote signal transmission.
[0063] Tap-c interface: Connect a data cable to charge the device, or connect a wired mobile camera to replace an ordinary camera, and automatically capture dangerous points in the scene image in a small or dim space. Specifically, the device can be connected to an external device such as Figure 3 The wired mobile camera shown in the figure, when successfully connected, directly jumps to the mobile camera's perspective on the display screen, and can capture images of the surrounding environment, thereby automatically capturing and marking dangerous points. The operator can debug the camera through the brightness adjustment button and clarity adjustment knob, and use the mobile camera to assist on-site operations.
[0064] To facilitate battery replacement and maintenance, the device includes a battery cover, which is screwed onto the back of the housing in the battery area. The cover encloses the battery, which contains a large-capacity 3200mAh lithium battery. In single-shot recording mode, the power consumption is ultra-low, with an 8-hour battery life. A 70mAh button cell battery is also built in for easy replacement of the backup lithium battery.
[0065] Finally, the device also includes a locator installed in the shell, which is connected to the cloud via a wireless communication module.
[0066] like Figure 4 As shown, the present invention provides an electronic safety warning method based on image recognition, which is applicable to any of the above-mentioned electronic safety warning devices based on image recognition, comprising: Step 410: After the device is turned on, the operator scans the QR code on the security pass worn by the operator through the camera to log in, and then selects the working mode after logging in; Step 420: After selecting the working mode, the camera takes real-time photos of the work site and transmits the captured images to the cloud via the wireless communication module; Step 430: The cloud identifies risk points on the received image using a pre-stored risk point identification algorithm. In step 440 , the cloud marks the identified risk points and transmits the images marked with the risk points back to the display screen on the device via the wireless transmission module for displaying warnings.
[0067] When in use, after pressing the power button, the device can interact with the operator, the master control platform and the cloud through the wireless communication module, and can complete manual or automatic switching or on-demand switching according to different scenes and geographical environments in real time to transmit data. Figure 5 The main menu is shown.
[0068] After the staff selects the device login function menu, enter Figure 6 On the device login screen shown, scan the QR code on the primary operator's security pass. After confirming the relevant information, you can log in to the current device. After logging in, the device will record the current operator's specific location and locate them in real time. During on-site operations, security risks can be identified using real-name identification.
[0069] After the staff selects the electronic map function menu, enter Figure 7 The electronic map interface shown shows the real-time location information of the staff so that the staff can check the walking route.
[0070] After the staff selects the public network intercom function menu, enter Figure 8 In the public network intercom interface shown, on-site personnel can remotely interact with main site personnel and online experts. Main site personnel and online experts can provide operational guidance and monitoring to on-site personnel, and conduct real-time and effective supervision and control of the safety organization measures, safety technical measures, and safety equipment at the work site. This process supports multi-person intercom and video interaction.
[0071] After the staff selects the video recording function menu, enter the following Figure 9 The image capture interface shown above allows for the capture and retrieval of multiple sets of images of workers, safety tools, and the work site. By automatically capturing static and dynamic images and comparing them against an electronic database, the device identifies potential hazards within these images. For static image capture, the device overlays and calculates multiple static images captured in the current work environment, analyzing risk points and generating hazard analysis data to remind workers to adhere to safety precautions. For dynamic image capture, the device monitors the entire work process using dynamic images captured in the current work environment, allowing back-end supervisors to remind workers to adhere to safety precautions. When a worker logs in, the device defaults to live dynamic image recording mode, which can be manually switched. This feature aims to regularize safety management requirements. By providing real-time monitoring from the worker's perspective, the device fosters safe working habits and prevents accidents resulting from violations of regulations, including personal injury, death, and power grid and equipment safety incidents.
[0072] Among them, the electronic safety fence setting interface in the video recording menu is as follows: Figure 10 As shown, the device captures static or dynamic images and compares them with data stored in the database that meets safety work standards. It analyzes and marks dangerous points on the work site, and simulates and calculates the safe working distance in real time based on the on-site working environment to ensure that the workers can carry out their work within a safe and reliable range. If the workers exceed the expected simulated safety distance, the device will issue an alarm to prompt the workers.
[0073] The video recording setting interface in the video recording menu is as follows: Figure 11 As shown, the system uses local cameras and wired mobile cameras to capture images based on the work environment, image resolution requirements, and whether to switch between night vision modes. Before setting up the camera, the interface prompts whether to enable thermal and infrared imaging. Enabling thermal imaging further identifies potential hazards within the workspace, such as bird nests, animals, and exposed wires. Enabling infrared imaging facilitates clearer imaging in dark conditions, improving image quality.
[0074] After the staff selects the remote intelligent connection function menu, enter Figure 12 The remote intelligent connection interface shown above allows you to upload live video or static images of on-site operations to the cloud for comparison with the backend database or to improve autonomous learning. Uploading to the cloud database: The device uploads specific hazardous point examples on site and completes the autonomous learning function by improving its own database of hazardous point examples.
[0075] After the staff selects the inspection and patrol function menu, they will enter Figure 13 The inspection and patrol menu interface shown ensures the stable operation of power equipment. The device conducts regular inspection, testing, maintenance and monitoring of power equipment to discover hidden dangers and faults of the equipment and perform timely repairs and maintenance to ensure the reliability and safety of the power system.
[0076] After the staff selects the image review function menu, enter Figure 14 The image review interface shown above allows viewing of all original dynamic and static images recorded by the device. For static images, the Risk Point Analysis button generates risk points in the current static image with a single click. Workers can then analyze the images. The static images captured by colleagues serve as a learning basis for the device's "smart assistant," enriching its electronic database. If similar work situations arise at subsequent work sites, the device can conduct intelligent analysis, summaries, and autonomous learning, providing timely preventative measures for staff. For dynamic images, workers or back-end monitoring can use image review to monitor the entire process and analyze risk points. Dynamic images provide support and analysis during on-site team work summaries. Image review can also assist staff in optimizing and correcting electronic fence settings to ensure they meet on-site requirements.
[0077] After the staff selects the smart assistant function menu, enter Figure 15 The smart assistant interface shown uses AI basic recognition functions. When staff hold a pre-shift meeting during on-site work, they can use AI functions to perform facial recognition and confirm work team members. During on-site work, AI functions can be used to identify the numbers of power lines, equipment, etc., and confirm the work location and objects.
[0078] After the staff selects the personal information function menu, enter Figure 16 The personal information interface shown is used to query the basic information of the operator, making it easier for back-end personnel to implement supervision.
[0079] After the staff selects the recording call function menu, enter Figure 17 The recorded call interface shown is used by operators to conduct real-time business communication with background monitoring personnel to ensure the smooth completion of on-site operations.
[0080] After the staff selects the message center function menu, enter Figure 18 The message center interface shown provides prompts for various information received by the device and effectively records each on-site operation, making it convenient for monitoring personnel to make subsequent inquiries.
[0081] After the staff selects the local setting function menu, enter Figure 19 The local setting interface shown can be used to set communication parameters, shooting parameters, recording parameters, brightness, danger point prompts and other functions.
[0082] Specifically, the pre-stored risk point identification algorithm adopts the SIFT algorithm. The SIFT algorithm compares the features of the received image with the risk point standard image pre-stored in the cloud database to confirm whether there is a risk point in the received image.
[0083] A further improvement of this technical solution is that the method for identifying risk points in the received image using the SIFT algorithm specifically includes: Apply Gaussian function to blur and downsample the image captured by the camera, construct a Gaussian pyramid, form a scale space, and detect and confirm several key point candidates in the scale space; Perform Taylor series expansion on each key point candidate to locate its position and scale; Remove low-contrast points and edge response points from the key point candidates to determine the key points; For each key point, a 16x16 neighborhood is taken around it with each key point as the center, and it is divided into 4x4 sub-blocks to determine the neighborhood sampling area; Sampling is performed in the determined neighborhood sampling area, and the gradient amplitude and gradient direction of each sampling point are calculated; Generate a direction histogram in the neighborhood sampling area of the key point based on the calculated gradient amplitude and gradient direction, and determine the main direction of the key point; Calculate the gradient histogram in 8 directions for each sub-block in the neighborhood sampling area to form an 8-bin histogram. For each key point, obtain a 128-dimensional feature vector to form a key point descriptor; The extracted key point descriptors are compared with the features of the risk point standard images stored in the cloud database.
[0084] Among them, the descriptor parameters are: sub-block size (such as 4x4), the number of bins of the direction histogram (such as 8 directions).
[0085] Furthermore, the Gaussian kernel function is: ; in, is the standard deviation of the Gaussian kernel, which controls the degree of blur.
[0086] In addition, the formula for calculating the gradient amplitude of the sampling point is: ; Among them, m(x, y) is the gradient amplitude of the sampling point, L(x+1, y) is the pixel value of the image at point (x+1, y), L(x-1, y) is the pixel value of the image at point (x-1, y), L(x, y+1) is the pixel value of the image at point (x, y+1), and L(x, y-1) is the pixel value of the image at point (x, y-1).
[0087] In addition, the formula for calculating the gradient direction of the sampling point is: ; in, is the gradient direction of the sampling point.
[0088] Although the present invention has been described in detail with reference to the accompanying drawings and in conjunction with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and substance of the present invention, persons of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions shall fall within the scope of the present invention. Any changes or substitutions that can be readily conceived by persons skilled in the art within the technical scope disclosed in the present invention shall fall within the scope of protection of the present invention.
Claims
1. An electronic safety warning device based on image recognition, characterized in that: The device comprises a housing and a magnetic back clip. The front of the housing is provided with a display screen, a buckle is provided above the display screen, function key areas are provided on the left and right sides of the housing, a camera is installed on the back of the housing, and infrared imaging lights are provided on the upper and lower sides of the camera. Inside the housing are a wireless communication module, an integrated relay module, a circuit board containing a controller, and a battery that powers the entire device. The buckle is connected to the magnetic back clip, and the magnetic back clip can be worn on the chest of a worker or attached to iron objects in the work area. The camera is connected to the input end of the controller, and the camera, infrared imaging light and display screen are all connected to the output end of the controller. The controller uploads the dynamic or static images taken by the camera to the cloud through the wireless communication module, identifies the risk points of the uploaded images through the risk point identification algorithm pre-stored in the cloud, and marks the identified risk points on the images. The images marked with risk points are transmitted to the display screen through the wireless communication module for display warnings.
2. The electronic safety warning device based on image recognition according to claim 1, characterized in that: It also includes a locator installed in the shell, which is connected to the cloud via a wireless communication module.
3. The electronic safety warning device based on image recognition according to claim 1, characterized in that: The function key area is provided with a thermal imaging start key and an infrared imaging light start key. Both the thermal imaging start key and the infrared imaging light start key are connected to the input end of the controller. The thermal imaging start key is used to control whether the camera is in the thermal imaging working mode, and the infrared imaging light start key is used to control whether the infrared imaging light is turned on.
4. The electronic safety warning device based on image recognition according to claim 1, characterized in that: It also includes a Tap-c interface for charging the battery or connecting an external mobile camera.
5. The electronic safety warning device based on image recognition according to claim 1, characterized in that: It also includes a Beidou communication start key, which is connected to the input end of the controller. The wireless communication module adopts the Beidou communication module, and the Beidou communication start key is used to control the opening and closing of the Beidou communication module.
6. An electronic safety warning method based on image recognition, applicable to the electronic safety warning device based on image recognition according to any one of claims 1 to 5, characterized in that: include: After the device is turned on, the operator logs in by scanning the QR code on the security pass worn by the operator through the camera, and then selects the working mode after logging in; After selecting the working mode, the camera will take real-time photos of the work site and transmit the captured images to the cloud via the wireless communication module; The cloud identifies risk points in the received images based on pre-stored risk point identification algorithms; The cloud marks the identified risk points and transmits the images marked with risk points back to the display screen on the device through the wireless transmission module for displaying warnings.
7. The electronic safety warning method based on image recognition according to claim 6, characterized in that: The pre-stored risk point identification algorithm uses the SIFT algorithm. The SIFT algorithm compares the features of the received image with the risk point standard image pre-stored in the cloud database to confirm whether there are risk points in the received image.
8. The electronic safety warning method based on image recognition according to claim 7, characterized in that: The method of identifying risk points on the received image using the SIFT algorithm specifically includes: Apply Gaussian function to blur and downsample the image captured by the camera, construct a Gaussian pyramid, form a scale space, and detect and confirm several key point candidates in the scale space; Perform Taylor series expansion on each key point candidate to locate its position and scale; Remove low-contrast points and edge response points from the key point candidates to determine the key points; For each key point, a 16x16 neighborhood is taken around it with each key point as the center, and it is divided into 4x4 sub-blocks to determine the neighborhood sampling area; Sampling is performed in the determined neighborhood sampling area, and the gradient amplitude and gradient direction of each sampling point are calculated; Generate a direction histogram in the neighborhood sampling area of the key point based on the calculated gradient amplitude and gradient direction, and determine the main direction of the key point; Calculate the gradient histogram in 8 directions for each sub-block in the neighborhood sampling area to form an 8-bin histogram. For each key point, obtain a 128-dimensional feature vector to form a key point descriptor; The extracted key point descriptors are compared with the features of the risk point standard images stored in the cloud database.
9. The electronic safety warning method based on image recognition according to claim 8, characterized in that: The formula for calculating the gradient amplitude of the sampling point is: ； Among them, m(x, y) is the gradient amplitude of the sampling point, L(x+1, y) is the pixel value of the image at point (x+1, y), L(x-1, y) is the pixel value of the image at point (x-1, y), L(x, y+1) is the pixel value of the image at point (x, y+1), and L(x, y-1) is the pixel value of the image at point (x, y-1).
10. The electronic safety warning method based on image recognition according to claim 9, characterized in that: The formula for calculating the gradient direction of the sampling point is: ； in, is the gradient direction of the sampling point.