A Field Electromagnetic Flowmeter On-site Environment Auxiliary Monitoring Method and System

Abstract

The present invention discloses a method and system for auxiliary monitoring of the on-site environment of a field electromagnetic flowmeter. The present invention includes an intelligent control cabinet; multiple groups of cameras radially installed; a sensor module for dedicated digital image processing, which is used to parse the video stream collected by the monitoring cameras, perform dynamic background accumulation, and generate real-time background frame images. By performing differential operations on the current sampled image and the background frame image, abnormal pixel points are captured and analyzed by artificial intelligence. Through human skeleton extraction, vehicle template recognition and matching, and intelligent license plate extraction, information about personnel, vehicles, and license plates in case of abnormalities at the scene is obtained, and the images that have been intelligently recognized and the attached recognition information are actively sent to the control cabinet. The present invention effectively improves the transmission distance of low-power terminal devices and large-range multi-point coverage through wireless transmission devices. Monitoring through machine vision saves the workload of manual inspection and improves the monitoring efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of field environment monitoring, and particularly relates to a method and system for assisting in monitoring the on-site environment of a field electromagnetic flowmeter. Background Art

[0002] Traditional video monitoring technology uses a monitoring camera and a control host, and the image acquisition and transmission are carried out by laying cables in the middle; however, the field working conditions are harsh and the transmission distance is very long. Laying cables is time-consuming, laborious, and consumes a large amount. Using 5G transmission technology can achieve wireless connection from the monitoring site to the control room. The 5G network adopts a point-to-point data transmission mode and uses high-performance and high-bandwidth wireless transmission devices, which can effectively increase the transmission distance of low-power terminal devices and large-scale multi-point coverage.

[0003] The latest digital image processing technology can perform intelligent monitoring, manual discrimination, real-time transmission, and remote alarm on the monitoring site. Integrating advanced digital image processing technology can subvert the previous operation mode of the manual duty control room. It can use machine vision to replace human eyes to distinguish abnormal monitoring situations, eliminating the work burden of manual 24-hour duty and on-site inspection. Summary of the Invention

[0004] In view of the above problems, the present invention proposes a method and system for assisting in monitoring the on-site environment of a field electromagnetic flowmeter.

[0005] One aspect of the present invention provides a method for assisting in monitoring the on-site environment of a field electromagnetic flowmeter, including:

[0006] Step s1. Real-time monitor the on-site environment of the electromagnetic flowmeter through multiple monitoring cameras, and dynamically accumulate the digital images sampled at intervals within a certain period of time to synthesize an original background image.

[0007] Step s2. Perform convolution operations on the sampled images obtained by each monitoring camera to synthesize a background frame image that is updated in real time dynamically.

[0008] Step s3. Convert the real-time obtained video frame image and the background frame image into grayscale, convert the color image into a grayscale image, and perform mean filtering and denoising on the grayscale image.

[0009] Step s4. Perform a difference operation on the video frame image and the background frame image that have undergone the above processing to obtain a foreground frame image to be detected.

[0010] Step s5. Perform adaptive binarization processing on the foreground image to obtain a binarized image. Then perform image geometric morphology processing on the binarized image and extract the edge contour.

[0011] Step s6. Locate the coordinates of the obtained edge contour and feedback the located coordinates to the video frame image.

[0012] Step s7. Extract the human skeleton from the coordinate location area of the video frame image to obtain the personnel target.

[0013] Step s8. Perform vehicle template recognition and matching on the coordinate location area of the video frame image.

[0014] Step s9. Obtain the license plate recognition result based on the recognized vehicle.

[0015] Step s10. Obtain the complete number of personnel, the number of vehicles, and the license plate numbers, and feedback the corresponding information to the intelligent control cabinet.

[0016] Step s11. The control cabinet collects the feedback information. Once each monitoring camera recognizes a person and approaches the electromagnetic flowmeter in a continuous video stream and stays for a long time, it enters the alarm state.

[0017] Save the current frame image in the alarm state, reproduce the alarm information on the current frame image in the form of a watermark, and perform line coding and encapsulation on it.

[0018] Call the digital image wireless transmission module for remote wireless transmission.

[0019] Step s12. After receiving the alarm information, the remote wireless personal terminal and the control room pop up a window and issue a warning to the monitoring personnel. The personnel can call the on-site buzzer generator to drive away the person on the site by clicking the alarm button.

[0020] Another aspect of the present invention provides a field electromagnetic flowmeter on-site environment auxiliary monitoring system, including:

[0021] An intelligent control cabinet, which is used to unidirectionally send working instructions to control each module, provides a physical medium for digital image storage, and integrates a picture sending server for remote transmission of digital images.

[0022] Multiple groups of cameras installed radially, and the paired monitoring cameras are used to achieve full-time and non-blind-spot monitoring of the remote monitoring site, forming a video link for the analysis of on-site situations.

[0023] A dedicated digital image processing sensor module, which is used to parse the video stream collected by the monitoring camera, perform dynamic background accumulation, and generate a real-time background frame image. By performing differential operations between the current sampled image and the background frame image, abnormal pixel points are captured and artificial intelligence analysis is performed. Through human skeleton extraction, vehicle template recognition and matching, and intelligent license plate extraction, information about personnel, vehicles, and license plates in case of abnormalities at the site is obtained, and the intelligently recognized image with the recognition information is actively sent to the control cabinet.

[0024] A digital image wireless transmission module is used for point-to-point wireless remote transmission of abnormal images sent from a sensor module to a control cabinet. The images will be sent to a control room and personal terminals to alert the staff in the remote monitoring room of abnormal situations on site.

[0025] A solar battery pack generates solar power in sunny weather to provide power for the monitoring site and store energy in a storage battery.

[0026] A pipeline mechanical transmission power supply and energy storage module is formed by connecting a small-diameter functional pipe section to a fluid pipeline. The two are connected by a ball valve, and the ball valve sends opening and closing instructions through a control cabinet. As the medium flows through the functional pipe section, it drives the mechanical rotor to rotate, converting the mechanical energy of the fluid medium into electrical energy, and transmitting the electrical energy to the storage battery through a transmission line for energy storage.

[0027] The beneficial effects of the present invention: Through the wireless transmission device, the present invention effectively improves the transmission distance and large-range multi-point coverage of low-power terminal devices. Monitoring through machine vision saves the workload of manual inspection and improves the monitoring efficiency. At the same time, through the combination of a solar panel and a pipeline mechanical transmission power supply and energy storage module, seamless connection of power consumption requirements throughout the day is achieved, effectively reducing the occurrence of faults at the monitoring site. Description of the Drawings

[0028] Figure 1 It is a structural diagram of a field electromagnetic flowmeter on-site environment auxiliary monitoring system based on machine vision. Detailed Embodiment

[0029] As Figure 1 shown, the system of this embodiment includes four groups of radially placed monitoring cameras 1, the electromagnetic flowmeter 2 to be monitored, a sensor module 3 for dedicated digital image processing, a digital image wireless transmission module 4, an intelligent control cabinet 5, a storage battery 6, a pipeline mechanical transmission power supply and energy storage module 7, and a solar battery pack 8.

[0030] The control cabinet in this embodiment contains a low-power control chip for unidirectionally sending working instructions to control each module. It provides a physical medium that can be used for digital image storage. It integrates a picture sending server that can be used for remote transmission of digital images. It can also implement an intelligent module for buzzer alarm according to algorithm requirements.

[0031] The paired monitoring cameras in this embodiment can achieve full-time domain and non-blind spot monitoring of the remote monitoring site, forming multiple video links for analysis of on-site situations. The camera can also turn on the night vision function to face the monitoring work at night or in complex weather conditions.

[0032] The sensor module for dedicated digital image processing in this embodiment can parse the video stream collected by the surveillance camera, perform dynamic background accumulation, and generate a real-time background frame image. By performing differential operations on the current sampled image and the background frame image, abnormal pixel points are captured and artificial intelligence analysis is carried out. Through human skeleton extraction, vehicle template recognition and matching, and intelligent license plate extraction, information about people, vehicles, and license plates in case of abnormalities at the scene is obtained; and the image that has been intelligently recognized is actively sent to the control cabinet along with the recognition information.

[0033] The digital image wireless transmission module in this embodiment can perform point-to-point wireless remote transmission on the abnormal images sent by the sensor module to the control cabinet. The images will be sent to the control room and personal terminals to alert the staff in the remote monitoring room that there are abnormalities at the scene.

[0034] The solar cell battery pack in this embodiment uses solar photovoltaic panels. The solar photovoltaic panels can generate solar power in sunny weather to provide power for the monitoring site and store energy in the storage battery.

[0035] The pipeline-type mechanical transmission power supply and energy storage module in this embodiment mainly considers that solar photovoltaic power generation cannot operate at night and in rainy and cloudy weather. A backup and stable power supply mode is required. This module connects a small-diameter functional pipe section to the fluid pipeline. The two are connected by a ball valve, and the ball valve sends opening and closing instructions through the control cabinet. When the medium flows through the functional pipe section, it drives the mechanical rotor to rotate, converting the mechanical energy of the fluid medium into electrical energy. And the electrical energy is transmitted to the storage battery through the transmission line for energy storage.

[0036] Based on the above system, in another embodiment, a method for assisting in monitoring the on-site environment of a field electromagnetic flowmeter is provided:

[0037] Step s1. The electromagnetic flowmeter on-site environment is monitored in real time through four-directionally installed surveillance cameras, and dynamic accumulation is performed on the digital images sampled at intervals within a certain period of time. A frame of original background image G0(x, y) is synthesized and stored in a specific directory. Specifically as follows:

[0038] G0(x, y) = (Im1(x, y) + Im2(x.y) + Im3(x, y) +... + Im n (x, y))

[0039] Step s2. Convolution operations are performed on the sampled images obtained by each surveillance camera to synthesize a frame of real-time dynamically updated background frame image G m+1 (x, y). Specifically as follows:

[0040] G m+1 (x, y) = G m(x, y) * deta + G m-1 (x, y) * (1 - deta)

[0041] where G m (x, y) is the current background frame image, and G m+1 (x, y) is the newly generated background frame image after convolution operation. deta is the integral operator of the convolution operation, and m takes an integer greater than 1, which can be specifically set separately in the system.

[0042] Step s3. Convert the video frame image and the background frame image obtained in real time into grayscale, convert the color image into a grayscale image, and perform mean filtering and denoising on the grayscale image.

[0043] Step s4. Perform a difference operation on the video frame image and the background frame image G m+1 (x, y) to obtain the foreground frame image to be detected. Specifically as follows:

[0044] cvAbsDiff(grayImg, filterImg, upImg);

[0045] where grayImg is the grayscale image, filterImg is the image after mean filtering, upImg is the foreground image after the difference operation, and cvAbsDiff is the difference operation function.

[0046] Step s5. Perform adaptive binarization processing on the foreground image to obtain a binarized image. Then perform image geometric morphology processing on the binarized image and extract the edge contours.

[0047] Step s6. Locate the coordinates of the obtained edge contours and feedback the located coordinates to the video frame image.

[0048] Step s7. Extract the human skeleton from the coordinate positioning area of the video frame image to obtain the personnel target. Mainly perform image segmentation on the pixel set in the positioning area, approximate the boundary of the pixel points that conform to the human body posture characteristics, then calculate through the coordinate domain where the pixel points are located and the characteristic pixels in the library, and perform pose matching according to the established models and the model libraries of standing, semi-squatting, and walking. Once the pose matching is successful, mark and count the personnel status in the target area.

[0049] The specific implementation method is as follows:

[0050] HumanDefine(DetectiingImg, body(x1, y1, x2, y2), Lineth, F)

[0051] Among them, HumanDefine is a function that accepts edge contour positioning, DetectiingImg is the current video frame image to be analyzed, body(x1,y1,x2,y2) is the coordinate area of the suspected human target after image segmentation, Lineth is the human target scale analysis parameter (this parameter can be preset by the system). F is the analysis result feedback parameter. If the analysis result is a suspected person, it is feedback as true, otherwise it is false.

[0052] Function calculation logic:

[0053]

[0054] Once the analysis result feedback parameter is feedback as true, then extract the image of the contour area and input it into the human pose recognition library in the system.

[0055] AnalysingImg = FindCountour(DetectiingImg,area)

[0056] Among them, AnalysingImg is the image of the extracted contour area, FindCountour is the function that performs contour area extraction, and area is the color image area and depth function to be extracted.

[0057] Tensor(AnalysingImg,softmax(ecopchs,alpha),instruct)

[0058] Among them, Tensor is the human recognition function, AnalysingImg is the function that inputs the human target contour area, softmax is the pooling function, ecopchs is the number of iterations, alpha is the learning rate, and instruct is the input interface of the human pose recognition library.

[0059] Select a suspected human area pixel of n1*n2 pixel size in AnalysingImg, input it into the softmax pooling function, and perform convolution operations with six feature maps of n1*n2 pixel size contained therein. Then connect the six convolution operation results with the feature maps of any 2*2 size area in the pose library. Perform ecopchs iterations at the alpha learning efficiency.

[0060] Calculate the convergence probability of the final iteration result:

[0061]

[0062] Among them, S(x) is the probability calculation formula, and x is the activation times after a series of processes. When the value of S(x) falls within a specific function interval and shows an upward trend, the function is determined to converge. The pixels of the suspected human body area are matched with the feature map, the human torso is marked, and the torso coordinates are obtained.

[0063] The marked torso coordinates are input into the function for human body structure geometric shape analysis.

[0064] DrawLine(AnalysingImg, E(Xn1,Yn1), F(Xn2,Yn2), NUM)

[0065] Among them, DrawLine is the drawing function, E(Xn1,Yn1), F(Xn2,Yn2) are the coordinates of a specific torso, and NUM is the coordinate order of a specific torso. Then, the geometric coordinate positions of each torso are matched with the preset pattern, the human body posture is analyzed, and the target counting is performed.

[0066] Step s8. Perform vehicle template recognition and matching on the coordinate positioning area of the video frame image. Specifically, the pixel area of the coordinate area is divided into intervals according to the gray frequency histogram, and the threshold parameters are set according to several clustering peaks. The edge contour of the vehicle is extracted according to the pixel values within the threshold interval, and the recognized contour is matched with the standby vehicle type library. Once the type matching is successful, the target area is marked and counted.

[0067] The specific implementation method is as follows:

[0068] Car_hist(AnalysingImg.ravel(), draw(0,256))

[0069] Among them, Car_hist is the gray frequency histogram drawing function, AnalysingImg.ravel() realizes the color-to-gray conversion of the original image and the traversal of the gray image. draw(0,256) realizes the statistics and drawing of the gray pixel distribution interval.

[0070] Because the color of the car tire is fixed and is in a specific interval in the frequency distribution histogram. The color of the surrounding car body is fixed. Once there are two clustering peaks, one large and one small, in the frequency distribution histogram, and one of the clustering peaks is within a certain specific frequency spectrum interval, the target is recognized as a vehicle. The vehicle edge contour is extracted.

[0071] Car_extract(AnalysingImg.contour(), expres)

[0072] Among them, Car_extract is the vehicle edge contour grabbing function, AnalysingImg.contour() is the contour drawing function, and expres is the difference threshold parameter between the vehicle color and the surrounding environment color. First, the initial contour of the vehicle is obtained, and then the color difference between the pixel points around the contour and the edge environment is compared. Once the difference exceeds the threshold parameter, the contour is contracted, and once the difference is less than the contour parameter, the contour is expanded.

[0073] Finally, the similarity between the solidified vehicle contour and the model contours of various types of engineering vehicles in the model library is calculated. Vehicles with a similarity higher than 50% are determined and counted.

[0074] Step s9. License plate recognition. Perform contour detection of the rectangular frame on the image area recognized as a vehicle. Once the specific area position is recognized, the area image is grayscale processed, then the grayscale image is character segmented, and the segmented characters are subjected to digital detection. For the unrecognized letter part, each character is binarized, the feature points of the processed binarized characters are extracted, and the structure of the feature points is input into the letter recognition library for one-by-one matching, and the successfully matched letters are fed back. Finally, the recognized letters and numbers are arranged in the original order to output the latest license plate string.

[0075] The specific implementation method is as follows:

[0076] cvRectangle(AnalysingImg - car, cvPoint, color)

[0077] Among them, cvRectangle is the rectangular frame recognition function, AnalysingImg - car is the target area recognized as a vehicle, cvPoint is the finally recognized coordinate parameter, and color is the color for drawing the rectangular frame.

[0078] cvColor(AnalysingImg - carNumber, gray, CV_BGR2GRAY)

[0079] Among them, AnalysingImg - carNumber is the license plate area image, and gray is the license plate grayscale image. CV_BGR2GRAY is the specific code segment to be executed.

[0080] Number_extract(gray, hierarchy, Num(n1, y))

[0081] Among them, Number_extract is a character segmentation function, gray is the grayscale image of the license plate, hierarchy is the called growth function, and Num(n1, y) is the intercepted digital image area and storage path. The Hierarchy function is called to take a growth point at the edge area of the upper left corner of the grayscale image, and traverse and extract the pixel values of the adjacent areas. For the pixel points with no obvious change in pixel value, annihilation processing is performed, and their pixel values are actively set to 255. The pixel points with obvious change in pixel value retain their original pixel values, and a new growth point is created. Taking the new growth point as a benchmark, compare the pixel values with the surrounding pixel points, mark the coordinate of the pixel area with no obvious change, and draw it in the n1 image. After the drawing is completed, store the image n1 into the file directory of the y path.

[0082] Read the extracted images one by one from their respective storage paths, and perform overlapping processing on the digital templates in the library and the characters in the images:

[0083] Number_read(y,ocr,static[0])

[0084] Among them, the Number_read function performs path reading, ocr is the read image, and static[0] is the digital template static library. When ocr completely overlaps with a certain number in the digital template, output the recognized number.

[0085] The images that cannot be overlapped are defaulted to Chinese characters and English characters.

[0086] Perform binarization processing on the images:

[0087] ThImg=Threshold(y,torch,60,cv2.THRESH_BINARY)

[0088] Among them, Threshold is the binarization function, torch is the read character image, 60 is the binarization threshold parameter, cv2.THRESH_BINARY is the code segment for performing binarization operation, and ThImg is the character image after binarization processing.

[0089] Extract the feature points of the binarized character image and compare it with the English character and Chinese character template libraries.

[0090] Word_read(ThImg,dynamic[0])

[0091] Among them, Word_read is the character recognition function, and dynamic[0] is the preset character template library.

[0092] Finally, arrange all the numbers and characters in the recognition order and splice and output them in the order of appearance.

[0093] Step s10. Obtain the complete number of personnel, the number of vehicles, and the license plate numbers. And feedback the corresponding information to the intelligent control cabinet.

[0094] Step s11. The control cabinet collects the feedback information. Once each monitoring camera recognizes a person and approaches the electromagnetic flowmeter in a continuous video stream and stays for a long time, it enters the alarm state. Save the current frame image in the alarm state, replicate the alarm information to the current frame image in the form of a watermark, and encode and package it. Call the digital image wireless transmission module 4 for remote wireless transmission.

[0095] Step s12. After the remote wireless personal terminal and the control room receive the alarm information, a pop-up window appears and a warning is issued to the monitoring personnel. The personnel can call the on-site buzzer generator to drive away the person on the site by clicking the alarm button.

Claims

1. A method for auxiliary monitoring of the on-site environment of a field electromagnetic flowmeter, characterized in that The method includes: Step S1. Use multiple monitoring cameras to monitor the on-site environment of the electromagnetic flowmeter in real time, and dynamically accumulate the digital images sampled at intervals within a certain period of time to synthesize an original background image; Step S2. Perform convolution operations on the sampled images obtained by each monitoring camera to synthesize a background frame image that is updated in real time and dynamically; Step S3. Perform gray-scale conversion on the video frame images obtained in real time and the background frame images to convert the color images into gray-scale images; and perform mean filtering and denoising on the gray-scale images; Step S4. Perform differential operations on the video frame images and the background frame images that have undergone the above processing to obtain the foreground frame images to be detected; Step S5. Perform adaptive binarization processing on the foreground images to obtain a binarized image; then perform image geometric morphology processing on the binarized image and extract the edge contours; Step S6. Locate the coordinates of the obtained edge contours and feedback the positioning coordinates to the video frame images; Step S7. Extract the human skeletons from the coordinate positioning areas of the video frame images to obtain personnel targets; Step S8. Perform vehicle template recognition and matching on the coordinate positioning areas of the video frame images; Step S9. Obtain the license plate recognition results based on the recognized vehicles; Step S10. Obtain the complete number of personnel, the number of vehicles, and the license plate numbers, and feedback the corresponding information to the intelligent control cabinet; Step S11. The control cabinet collects the feedback information. Once each monitoring camera recognizes a person and approaches the electromagnetic flowmeter in a continuous video stream and stays for a long time, it enters the alarm state; Save the current frame image in the alarm state, replicate the alarm information to the current frame image in the form of a watermark, and perform encoding and encapsulation on it; Call the digital image wireless transmission module for remote wireless transmission; Step S12. After the remote wireless personal terminal and the control room receive the alarm information, they pop up a window and issue a warning to the monitoring personnel. The personnel call the on-site buzzer generator to drive away the person on the site by clicking the alarm button; Among them, step S8 is specifically: Divide the pixel area of the coordinate area according to the gray-scale frequency histogram and set the threshold parameters according to several clustering peaks; Extract the edge contours of the vehicle according to the pixel values within the threshold interval, match the recognized contours with the standby vehicle type library. Once the type match is successful, mark and count the target area; Among them, step S9 is specifically: Perform contour detection of a rectangular frame on the image area recognized as a vehicle. Once the specific area position is recognized, perform gray-scale processing on the area image; Perform character segmentation on the gray-scale image, and perform letter and number detection on the segmented characters; Arrange the recognized letters and numbers in the original order and output the latest license plate string; For the unrecognized letter part, perform binarization processing on a single character, extract the feature points of the processed binarized character, and input the structure of the feature points into the letter recognition library for one-by-one matching, and feedback the successfully matched letters.

2. The monitoring method according to claim 1, characterized in that, Step S7 is specifically: Perform image segmentation on the pixel set in the positioning area, and approximate the boundary of the pixel points that conform to the human body posture characteristics; Calculate using the coordinate domain where the pixel points are located and the characteristic pixels in the library, and perform pose matching according to the established model and the model libraries of standing, semi-squatting, and walking. Once the pose matching is successful, mark and count the personnel status in the target area.

3. A field electromagnetic flowmeter on-site environment auxiliary monitoring system for implementing the method described in any one of claims 1-2, characterized in that, Including: An intelligent control cabinet, used to unidirectionally send working instructions to control each module, provide a physical medium for storing digital images; integrated with a picture sending server for remotely transmitting digital images. Multiple groups of radially installed cameras. The paired monitoring cameras are used to achieve full-time and non-blind-spot monitoring of the remote monitoring site, forming a video link for analyzing the on-site situation. A dedicated digital image processing sensor module, used to parse the video stream collected by the monitoring camera, perform dynamic background accumulation, and generate a real-time background frame image; capture abnormal pixel points by performing differential operations on the current sampled image and the background frame image and conduct artificial intelligence analysis; obtain information about personnel, vehicles, and license plates in case of abnormalities on-site through human skeleton extraction, vehicle template recognition and matching, and intelligent license plate extraction, and actively send the intelligently recognized image with the recognition information to the control cabinet. A digital image wireless transmission module, used to perform point-to-point wireless remote transmission of the abnormal images sent by the sensor module to the control cabinet. The images will be sent to the control room and personal terminals to alert the staff in the remote monitoring room that there are abnormalities on-site. A solar battery pack: Generate solar power on sunny days to provide power for the monitoring site and store energy in the battery. A pipeline-type mechanical transmission power supply and energy storage module, by connecting a small-diameter functional pipe section in the fluid pipeline; the two are connected by a ball valve, and the ball valve sends opening and closing instructions through the control cabinet; drive the mechanical rotor to rotate by the medium flowing through the functional pipe section, convert the mechanical energy of the fluid medium into electrical energy, and transmit the electrical energy to the battery for energy storage through the transmission line.

4. An on-site environment auxiliary monitoring system for a field electromagnetic flowmeter according to claim 3, characterized in that, When the weather is moderately sunny, use the solar battery pack to charge the battery module; in case of sudden complex weather conditions, when the energy storage of the battery reaches the low critical state, open the ball valve to allow the fluid medium in the pipeline to pass through the pipeline-type mechanical transmission power supply and energy storage module at a constant flow rate, driving the internal mechanical rotor generator of the module to generate electricity, and storing the electrical energy in the battery module through the line.

Images