A method for remote monitoring of a converter station based on image recognition

Abstract

The application discloses a kind of based on image recognition's converter station remote monitoring method.The present application includes the following steps: setting up camera to shoot the image of all physical assets located in converter station, the position information of physical asset is marked;The physical asset that needs to be checked is extracted from the database The image photographed by camera is carried out image recognition to data center;Setting up millimeter wave radar to detect the vital signs of personnel;Image recognition is carried out to determine whether smoke appears;If smoke appears, use millimeter wave radar to determine whether there is the vital signs of personnel in smoke;The electronic fence is constructed to the setting work area of converter station, then image recognition is carried out, and personnel, vehicle entering and exiting setting work area are identified.The present application can remotely check the physical assets in converter station, improve the efficiency and timeliness of checking.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of converter station monitoring system, and particularly relates to a remote monitoring method of converter station based on image recognition. BACKGROUND

[0002] The converter station is a station site for converting AC power into DC power or converting DC power into AC power in a high-voltage direct current power transmission system. The inventory in the converter station refers to the inventory of various physical assets, and the re-inventory is the re-inventory of assets by different personnel from the inventory.

[0003] In the traditional inventory of the converter station, the relevant staff needs to go to the site for inventory. Although the on-site inventory has a wide range of adaptability, most of the property and materials can be taken in this way, but the workload is large, the efficiency is not high, and the inventory result feedback is not timely.

[0004] In the prior art, the monitoring system of the converter station, for example, the Chinese patent with the patent number 201510661510.8 discloses a panoramic state monitoring system of the converter station, which comprises a data collection module, a data storage module, a data analysis processing module and an application module. The monitoring data of various devices is effectively fused and correlated, and the panoramic monitoring and multi-dimensional analysis of the device state in the ultra-high voltage converter station are realized. However, the prior art does not include the inventory of physical assets, and does not include the monitoring and inventory of whether the physical assets and safety signs are moved in violation of the rules. At the same time, it lacks the identification of personnel and vehicles entering the converter station, and it is difficult to ensure that the personnel entering the working area are authorized and permitted. SUMMARY

[0005] In order to solve one or more defects or problems in the prior art, the purpose of the present application is to provide a remote monitoring method of the converter station based on image recognition, so as to overcome the problems of manual inventory and the defects of the lack of inventory function of physical assets in the existing monitoring system.

[0006] The technical scheme of the present application is as follows:

[0007] A remote monitoring method of the converter station based on image recognition, comprising the following steps:

[0008] A plurality of cameras connected with a data center are arranged in the converter station, the cameras shoot images of all physical assets located in the converter station, the images of the physical assets are stored in the database of the data center, and the position information of the physical assets is marked in the database;

[0009] A safety sign is set when a person operates a physical asset, and then a personnel operation image captured by a camera is extracted from a database to a data center for image recognition, and a mechanical split state and an electrical indication state of the corresponding physical asset after personnel operation are obtained through image recognition, so as to realize inventory of the corresponding physical asset;

[0010] A plurality of millimeter wave radars connected with the data center are arranged in the converter station, and data of the millimeter wave radars are stored in the database of the data center, and the millimeter wave radars are used to detect vital signs of the personnel;

[0011] The image captured by the camera is extracted from the database to the data center for image recognition to determine whether smoke appears; if smoke appears, the millimeter wave radars are used to detect the smoke, vital signs of the personnel in the smoke are determined according to the data of the millimeter wave radars, and the data center issues an alarm if the vital signs of the personnel appear in the smoke;

[0012] In the data center, an electronic fence is constructed for a set work area of the converter station by using images of the entire converter station, and then images captured by the camera are extracted from the database for image recognition, personnel and vehicles entering and leaving the set work area are identified, and entering and leaving station data are established in the database.

[0013] Preferably, the inventory is realized by image recognition when the personnel operate the physical asset, and the specific steps include:

[0014] Each person entering the set work area in the converter station changes a work uniform;

[0015] In the set work area, the operated physical asset and the safety sign are placed in a view range of the camera and position coordinates of the physical asset and the safety sign are marked;

[0016] After the image captured by the camera is transmitted to the data center, the data center obtains the number of personnel entering the set work area and body contours of the personnel by image recognition of the work uniform;

[0017] When the personnel operate the physical asset, the data center identifies behaviors of the personnel according to the image captured by the camera, and determines whether a behavior of violating operation of the corresponding physical asset appears;

[0018] The data center records whether the behavior of violating operation appears, and records a mechanical split state and an electrical indication state of the physical asset after personnel operation by image recognition.

[0019] Further, the specific steps of placing the operated physical asset and the safety sign in the view range of the camera and marking the position coordinates of the physical asset and the safety sign include:

[0020] Adjust the position of the camera so that the real assets and safety signs are within the field of view of the identification camera;

[0021] The data center takes the lower left corner of the camera's viewfinder frame as the origin to establish a coordinate system, then calculates the position coordinates of the real assets and safety signs in the coordinate system, and then marks the position coordinates of the corresponding real assets and safety signs in the image.

[0022] Further, the image recognition of the work clothes obtains the number of personnel entering the set work area and the human body contour of the personnel, including the following steps:

[0023] The data center performs image recognition, and when a work clothes is recognized, the number of work clothes in the set work area is obtained according to the recognition;

[0024] The data center updates the background of the image to obtain a human body foreground image, and removes holes and white points in the human body foreground image;

[0025] The distance between the left and right points in the human body foreground image is set as the width of the identification frame of the human body contour, and the distance between the upper and lower points in the human body foreground image is set as the height of the identification frame of the human body contour, thereby obtaining the identification frame of the human body contour.

[0026] Further, the judgment of whether there is a violation operation corresponding to the behavior of the real asset and the safety sign is as follows:

[0027] Image recognition is performed in the data center to calculate whether the position coordinates of the real asset and / or safety sign change;

[0028] If the position coordinates of the real asset and / or safety sign change, the position coordinates of the center point of the human body contour identification frame in the set work area are calculated, and then the distance between the position coordinates of the center point of the human body contour identification frame and the position coordinates of the real asset and / or safety sign is calculated.

[0029] If the distance between the position coordinates of the center point of the human body contour identification frame and the position coordinates of the real asset and / or safety sign is less than a set threshold, it indicates that there is a behavior of personnel violating the operation of the corresponding real asset and / or safety sign.

[0030] Further, the data center sends an alarm message to the personnel in the set work area after identifying the behavior of violating the operation in the set work area.

[0031] Further, the data center detects smoke through image recognition, and the specific steps are as follows:

[0032] The image captured by the camera is processed using a Gaussian model to obtain the motion area of the image;

[0033] The image photographed by the camera is processed by using the dark channel defogging algorithm to obtain a smoke-free image model;

[0034] The image processed by the Gaussian model and the image processed by the dark channel defogging algorithm are subjected to difference processing to obtain a difference image; the difference image is subjected to binaryzation processing to obtain a smoke area to be identified in the image, and an intersection area between the smoke area and the motion area in the image is obtained;

[0035] A smoke classification model trained by deep learning is used to identify the smoke area to be identified in the image, and it is determined whether smoke appears in the image.

[0036] Further, after the data center detects the vital signs of the personnel in the smoke in the set work area using the data of the millimeter wave radar, it is determined whether the vital signs of the personnel are higher than a set value, and if the vital signs are lower than the set value, a rescue message is sent to other personnel.

[0037] Further, after the electronic fence is constructed, image recognition is performed on personnel entering and leaving the set work area, and the specific steps are as follows:

[0038] A camera matched with the height of the personnel is set, and the face image of the personnel entering and leaving the station is captured and transmitted to the data center;

[0039] The data center prewrites the face image and identity data of the personnel with the permission to enter and leave the converter station in the database;

[0040] The data center compares the captured face image of the personnel with the face image of the personnel with the permission to enter and leave the converter station in the database, marks whether the captured personnel have the permission to enter and leave the converter station in the captured image, and counts the number of personnel.

[0041] Further, after the electronic fence is constructed, image recognition is performed on vehicles entering and leaving the set work area, and the specific steps are as follows:

[0042] A camera matched with the height of the license plate of the vehicle is set, and the license plate image of the vehicle entering and leaving the station is captured and transmitted to the data center;

[0043] The data center converts the captured license plate image into a gray-scale image and removes noise;

[0044] The outline of the gray-scale image is extracted by using a sobel operator to obtain a gray-scale outline image, and the characters in the gray-scale outline image are cut out according to the projection of the image gray-scale value in the horizontal direction and the vertical direction according to the ratio of 1:2.5;

[0045] The license plate number is identified from the cut-out characters by using a template matching algorithm.

[0046] Compared with the prior art, the present application has the beneficial effects that:

[0047] By means of image recognition, the real assets in the converter station can be remotely inventoried; by calculating the position coordinates of the real assets, safety signs and human body contours, it is determined whether someone has moved the real assets and safety signs without authorization, and an alarm is given when the position or state of the real assets and safety signs is wrong, so as to realize efficient remote inventory, correct irregular operation in time and avoid the problems of low efficiency of manual inventory and untimely feedback of inventory results; through the camera and the millimeter wave radar, it can be determined whether someone is trapped in smoke when smoke fire occurs in the set working area, so as to command rescue; by identifying the personnel and vehicle information in the converter station, it is determined whether unauthorized personnel and vehicles enter the set working area, so as to ensure the safety of the set working area. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0049] Figure 1 is a general flowchart of a converter station remote monitoring method based on image recognition. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments thereof. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0051] EMBODIMENT

[0052] As shown in Figure 1 the converter station remote monitoring method based on image recognition of the present embodiment includes the following steps:

[0053] S1, a plurality of cameras connected with a data center are arranged in the converter station, the cameras shoot images of all real assets located in the converter station, the images of the real assets are stored in the database of the data center, and the position information of the real assets is marked in the database; wherein the real assets include devices such as knife switch, ground knife, switch, etc.

[0054] S2, setting a safety sign when the personnel operate the real assets, then extracting the personnel operation image shot by the camera from the database to the data center for image recognition, obtaining the mechanical split state and electrical indication state of the corresponding real assets after the personnel operation through image recognition, thereby realizing the inventory of the corresponding real assets;

[0055] The inventory of the converter station refers to the real-time monitoring of a series of operation processes of related switches, ground switches, switches and other devices during the power-off and power-on processes of the converter station, and the post-operation review; the power-off operation may involve the operation from the running to the hot standby state, the operation from the hot standby to the cold standby state, and the operation from the cold standby to the maintenance state; the power-on operation may involve the operation from the maintenance to the cold standby state, the operation from the cold standby to the hot standby state, and the operation from the hot standby to the running state;

[0056] In the embodiment, the specific steps of the inventory by image recognition when the personnel operate the real assets preferably include:

[0057] S21, making each personnel entering the set working area in the converter station change the work clothes;

[0058] S22, placing the real assets to be operated and the safety sign in the set working area in the view range of the camera and marking the position coordinates of the real assets and the safety sign; specifically including:

[0059] Adjusting the position or angle of the camera so that the real assets and the safety sign are located in the view range of the recognition camera; the camera spacing is about 6 meters when the lens frame size is 4mm and the irradiation angle is 70 degrees; the camera spacing is 6-10 meters when the lens frame size is 6mm and the irradiation angle is 50 degrees; the camera spacing is 10-20 meters when the lens frame size is 8mm and the irradiation angle is 38 degrees; the data center takes the lower left corner of the view frame of the camera as the origin to establish a coordinate system, then calculates the position coordinates of the real assets and the safety sign in the coordinate system, and then marks the position coordinates of the corresponding real assets and the safety sign in the image;

[0060] S23, after the image shot by the camera is transmitted to the data center, the data center obtains the number of personnel entering the set working area and the human body contour of the personnel through image recognition of the work clothes; specifically including:

[0061] When the work clothes are recognized, the data center adds one to the number of personnel in the set working area;

[0062] The data center models and updates the background of the image, and subtracts the current frame from the background to obtain the human body foreground image, and removes the holes and white points in the human body foreground image;

[0063] The distance between the left and right points in the human body foreground image is set as the width of the recognition frame of the human body contour, and the distance between the upper and lower points in the human body foreground image is set as the height of the recognition frame of the human body contour, so as to obtain the recognition frame of the human body contour;

[0064] S24, when the personnel operate the real assets, the data center identifies and analyzes the behavior of the personnel according to the image captured by the camera, and judges whether the behavior of violating the operation of the corresponding real asset appears; specifically including:

[0065] In the data center, the position coordinates of the real assets and / or safety signs are calculated to determine whether the position coordinates change; if the position coordinates of the real assets and / or safety signs change, the position coordinates of the center point of the human body contour recognition frame in the set working area are calculated, and then the distance between the position coordinates of the center point of the human body contour recognition frame and the position coordinates of the real assets and / or safety signs is calculated; if the distance between the position coordinates of the center point of the human body contour recognition frame and the position coordinates of the real assets and / or safety signs is less than a set threshold, it indicates that the personnel have the behavior of violating the operation of the corresponding real assets and / or safety signs;

[0066] S25, the data center records whether the behavior of violating the operation appears, and records the mechanical opening and closing state and the electrical indication state of the real assets after the personnel operate the real assets; after the data center identifies that the set working area has the behavior of violating the operation, the data center sends an alarm message to the personnel in the set working area, specifically an audible and visual alarm signal;

[0067] S3, a plurality of millimeter wave radars connected with the data center are arranged in the converter station, and the data of the millimeter wave radars are stored in the database of the data center, and the millimeter wave radars are used to detect the vital signs of the personnel; the millimeter wave radar has a very high working frequency and a short wavelength, and can have a high resolution, and can still be used normally in a smoke environment; the millimeter wave radar emits electromagnetic waves with a specific waveform through an antenna, and the electromagnetic waves are intercepted by a target in an effective radiation range, a part of the energy of the reflected electromagnetic waves is returned to the antenna and is received by the radar, and through amplification, signal processing and other processes, the position, moving speed, direction and other information of the target relative to the radar are finally calculated, so that the data center can calculate the vital signs of the personnel target by using the information;

[0068] S4, the image captured by the camera is extracted from the database to the data center for image recognition to determine whether smoke appears; if smoke appears, the millimeter wave radar is used to detect the smoke, and the data of the millimeter wave radar are used to determine whether there is a vital sign of personnel in the smoke; if there is a vital sign of personnel in the smoke, the data center sends an alarm; specifically including:

[0069] S41, the data center detects smoke through image recognition; the image captured by the camera is processed using a Gaussian model to obtain the motion area of the image; the image captured by the camera is processed using a dark channel defogging algorithm to obtain a smoke-free image model; the image processed by the Gaussian model and the image processed by the dark channel defogging algorithm are subjected to difference processing to obtain a difference image; the difference image is subjected to binaryzation processing to obtain the smoke area to be identified in the image, and the intersection area between the smoke area and the motion area of the image is obtained; a smoke classification model trained through deep learning is used to identify the smoke area to be identified in the image to determine whether smoke appears in the image; if smoke appears, the data center sends a message to notify personnel in the set smoke area to evacuate;

[0070] S42, after the data center detects that there is a person's vital sign in the smoke in the set work area using the data of the millimeter wave radar, it is determined whether the person's vital sign is higher than a set value; if it is lower than the set value, it means that the person's vital sign is unstable and dangerous, and then a rescue message is sent to other personnel to inform other personnel to rescue the person whose vital sign is unstable;

[0071] S5, in the data center, an electronic fence is constructed for the set work area of the converter station using the images of the entire converter station, and then images captured by the camera are extracted from the database for image recognition to identify personnel and vehicles entering and leaving the set work area, and data of entering and leaving the station are established in the database; specifically including:

[0072] S51, after the electronic fence is constructed, image recognition is performed on personnel entering and leaving the set work area, a camera matching the height of the personnel is set, the facial image of the personnel entering and leaving the station is captured and transmitted to the data center; the data center prewrites the facial image and identity data of personnel with the permission to enter and leave the converter station in the database; the data center compares the captured facial image with the facial image of personnel with the permission to enter and leave the converter station in the database to mark whether the captured personnel have the permission to enter and leave the converter station in the captured image, and counts the number of personnel;

[0073] S52, after the electronic fence is constructed, image recognition is performed on vehicles entering and leaving the set work area, a camera matching the height of the license plate of the vehicle is set, the license plate image of the vehicle entering and leaving the station is captured and transmitted to the data center; the data center converts the captured license plate image into a gray image and removes noise; the sobel operator is used to extract the contour of the gray image to obtain a gray contour image, and the area suspected of being a non-license plate number in the gray contour image is removed according to the ratio of 1:2.5 of length to width, and then the characters in the gray contour image are cut out according to the projection of the image gray value in the horizontal direction and the vertical direction; the license plate number is identified from the cut-out characters through a template matching algorithm.

[0074] Compared with the prior art, the embodiment has the beneficial effects that:

[0075] By means of image recognition, the real assets, safety signs and position information in the converter station can be remotely inventoried; by calculating the position coordinates of the real assets and safety signs and the position coordinates of the personnel body contour recognition frame, it is determined whether personnel have moved the real assets or safety signs without authorization, so as to prevent the real assets or safety signs from being disordered in position, realize accurate remote inventorying, prevent personnel from moving the related equipment without authorization and causing the loss of the equipment, realize efficient remote inventorying, timely correct the irregular operation behavior, avoid the problems of low efficiency of manual inventorying and untimely feedback of inventorying results; by means of detection of the camera combined with the millimeter wave radar, it can be determined whether personnel are trapped in smoke when smoke fire occurs in the set working area, and rescue is timely notified; by inputting the information of personnel and vehicles entering the converter station, it is prevented that personnel without authority enter the working area.

[0076] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A converter station remote monitoring method based on image recognition, characterized in that: The steps are as follows: multiple cameras connected to the data center are installed in the converter station, the cameras capture images of all physical assets located in the converter station, store the images of the physical assets in a database of the data center, and mark the location information of the physical assets in the database; When personnel operate physical assets, safety signs are set up. Then, images of personnel operation captured by cameras are extracted from the database and sent to the data center for image recognition. Through image recognition, the mechanical opening and closing status and electrical indication status of the corresponding physical assets after the personnel operation are obtained, thereby realizing the inventory of the corresponding physical assets. Multiple millimeter-wave radars connected to the data center are installed in the converter station. The data from the millimeter-wave radars is stored in the database of the data center. The millimeter-wave radars are used to detect the vital signs of personnel. Images captured by the camera are extracted from the database and sent to the data center for image recognition to determine whether smoke is present. If smoke is present, millimeter-wave radar is used to detect the smoke and determine whether there are any vital signs of people in the smoke based on the millimeter-wave radar data. If there are signs of life in the smoke, the data center will issue an alarm; Using images of the entire converter station in the data center, an electronic fence is constructed for the designated work area of ​​the converter station. Images captured by the cameras are then extracted from the database for image recognition. This identifies people and vehicles entering and exiting the designated work area, and creates entry and exit data in the database. When personnel operate physical assets, they use image recognition to perform inventory. The specific steps include: Have everyone entering the designated work area in the converter station change into work clothes; In the designated work area, place the physical assets and safety signs to be operated within the camera's field of view and mark their location coordinates; After the images captured by the camera are transmitted to the data center, the data center performs image recognition on the work clothes to obtain the number of people entering the set work area and their body outlines; When personnel operate physical assets, the data center identifies the personnel's behavior based on the images captured by the camera and determines whether there is any illegal operation of the corresponding physical assets; The data center records any violations and uses image recognition to identify the mechanical separation and connection status and electrical indication status of physical assets after human operation and records them. Place the physical assets and safety signs to be operated within the camera's field of view and mark their location coordinates. The specific steps include: Adjust the camera position so that the physical assets and safety signs are within the camera's field of view; The data center establishes a coordinate system using the lower left corner of the camera's viewfinder as the origin. It then calculates the coordinates of the physical assets and safety signs in the image within the coordinate system and marks the corresponding physical assets and safety signs in the image. Performing image recognition on work clothes to obtain the number of people entering the set work area and their body contours includes the following steps: The data center performs image recognition and when work clothes are identified, the number of work clothes in the work area is set according to the recognition; The data center updates the background of the image to obtain the human foreground image, and removes the holes and white spots in the human foreground image. The distance between the left and right points in the human foreground image is set as the width of the recognition frame of the human body contour, and the distance between the upper and lower points in the human foreground image is set as the height of the recognition frame of the human body contour, thereby obtaining the recognition frame of the human body contour; Determine whether there is any illegal operation of the corresponding physical assets and safety signs. The specific steps are as follows: Perform image recognition in the data center to calculate whether the location coordinates of physical assets and / or security signs have changed; If the position coordinates of the physical asset and / or the safety sign change, the position coordinates of the center point of the human outline recognition frame within the set working area are calculated, and then the distance between the position coordinates of the center point of the human outline recognition frame and the position coordinates of the physical asset and / or the safety sign is calculated; If the distance between the position coordinates of the center point of the human outline recognition frame and the position coordinates of the physical asset and / or safety sign is less than the set threshold, it means that a person has violated the regulations in operating the corresponding physical asset and / or safety sign.

2. The converter station remote monitoring method based on image recognition according to claim 1, characterized in that: After the data center identifies any illegal operations in the designated work area, it will send an alarm message to the personnel in the designated work area.

3. The converter station remote monitoring method based on image recognition according to claim 1, characterized in that: The data center detects smoke through image recognition. The specific steps are as follows: The image captured by the camera is processed using the Gaussian model to obtain the motion area of ​​the image; The image captured by the camera is processed using the dark channel defogging algorithm to obtain a smoke-free image model; Perform difference processing on the image processed by the Gaussian model and the image processed by the dark channel defogging algorithm to obtain a difference image; perform binarization processing on the difference image to obtain the smoke area to be identified in the image, and obtain the intersection area between the smoke area and the motion area of ​​the image; Use the smoke classification model trained through deep learning to identify the smoke area to be identified in the image and determine whether smoke appears in the image.

4. The method for remote monitoring of converter stations based on image recognition according to claim 3, characterized in that: The data center uses millimeter-wave radar data to detect the presence of human vital signs in the smoke in the set work area, and then determines whether the vital signs of the person are higher than the set value. If they are lower than the set value, a rescue message is sent to other personnel.

5. The converter station remote monitoring method based on image recognition according to claim 4 is characterized in that: After building the electronic fence, perform image recognition on people entering and leaving the set work area. The specific steps are as follows: Set up cameras that match the height of people to capture facial images of people entering and leaving the station and transmit them to the data center; The data center pre-writes facial images and identity data of personnel with access rights to the converter station into the database; The data center will compare the captured facial images of the people with the facial images of people with permission to enter and exit the converter station in the database, mark in the captured images whether the captured people have permission to enter and exit the converter station, and count the number of people.

6. The method for remote monitoring of converter stations based on image recognition according to claim 5, characterized in that: After building the electronic fence, perform image recognition on vehicles entering and leaving the set work area. The specific steps are as follows: Set up a camera that matches the height of the vehicle's license plate to capture images of the license plates of vehicles entering and leaving the station and transmit them to the data center; The data center converts the captured license plate image into a grayscale image and removes noise; Use the Sobel operator to extract the contour of the grayscale image to obtain a grayscale contour image, and then cut out the characters in the grayscale contour image according to the horizontal and vertical projections of the image grayscale values ​​with an aspect ratio of 1:2.5; The license plate number is identified from the cut characters through the template matching algorithm.

Images