A method and device for floating disc safety monitoring based on machine vision

Abstract

The present invention discloses a floating disc safety monitoring method and device based on machine vision, which can automatically monitor the leakage status of each module floating box all-weather, measure the height of the floating disc through the system to obtain the storage capacity of the tank, and measure the overall inclination degree of the floating disc; the advantages of the monitoring system are as follows: there is no circuit system inside the tank, and the vision and computing elements are arranged outside the tank, which has high safety; since the number of floating boxes of the floating disc is large, the leakage of a small number of floating boxes does not affect the safe operation, but after the leakage quantity accumulates, it will damage the operation safety. This system can know the leakage situation before the floating disc inclines, greatly improving the operation safety of the fully liquid-contact internal floating disc. During shutdown maintenance, the leaked floating boxes can also be replaced according to the monitoring results, greatly reducing the maintenance cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of floating roof monitoring for vertical storage tanks, and particularly relates to a floating roof safety monitoring method and device based on machine vision. Background Art

[0002] During the long-term storage of petroleum and its light derivatives in vertical storage tanks, there is an evaporation phenomenon, which will cause oil loss and pollution. The floating roof can significantly reduce the evaporation of oil products, so it is widely used at present. During the use of the floating roof, its safety is very important. Once an accident occurs in the storage tank, it will not only cause economic losses, but also damage personal safety and cause irreparable damage to the natural environment. If the floating roof leaks and loses buoyancy, a sunken roof phenomenon will occur, and the function of reducing evaporation will be lost. During the process of oil inlet and outlet, the floating roof needs to rise and fall with the change of the liquid level. If the floating roof leaks and causes tilting, a stuck roof phenomenon will occur, which will affect the oil inlet and outlet, and in severe cases, it will affect the overall safety of the storage tank. Therefore, the safety monitoring system of the floating roof is particularly important, but there are still many deficiencies in the current floating roof safety detection technology.

[0003] Chemicals derived from light petroleum such as gasoline will volatilize from the storage tank due to the phenomena of small breathing and large breathing during storage, which not only causes the loss of oil products, but also pollutes the environment and makes the storage plant area dangerous. Therefore, fully liquid-contact internal floating roofs are widely used in the industry to reduce the volatilization phenomenon. The fully liquid-contact internal floating roof is composed of many hollow floating boxes, so it can float on the oil. The operating state of the floating roof is closely related to the leakage condition of the floating boxes; however, the existing technology can only monitor the overall state of the floating roof, and the safety state of each floating box cannot be effectively monitored.

[0004] According to the existing technology, a floating roof status indicator for an external floating roof storage tank (application number 201410498716.9) is disclosed on the website of the State Intellectual Property Office. This invention patent uses a solar cell to drive a variety of sensors to monitor the floating roof and send signals. This invention patent has the following disadvantages. Although a waterproof and dustproof shell is designed, due to its use of electricity to drive, there is still a considerable risk when used in the storage environment of oil products.

[0005] According to the existing technology, an internal floating roof tank floating roof pose monitoring system and monitoring method (application number 201910214000.4) is disclosed on the website of the State Intellectual Property Office. This invention patent has the following disadvantages. Using multiple radar level gauges to monitor the floating roof attitude, only the overall attitude data of the floating roof can be obtained, and the leakage state of specific floating roof units cannot be measured.

[0006] According to the prior art, a floating roof inspection and monitoring device in an oil tank (application number CN202222955788.4) is disclosed on the website of the National Intellectual Property Administration. This utility model has the following defects: An observation window is opened on the side of the oil tank for manual monitoring, which has low accuracy and high cost and cannot monitor all day long. Summary of the Invention

[0007] The purpose of the present invention is to design a floating roof safety monitoring method and device based on machine vision, which can automatically monitor the leakage state of each module floating tank all day long, can also measure the height of the floating roof through this system to obtain the storage capacity of the tank, and can also measure the overall inclination degree of the floating roof.

[0008] In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0009] A floating roof safety monitoring method for vertical storage tanks based on machine vision, characterized in that the method is as follows:

[0010] S1: The supplementary light provides illumination for the floating roof in the storage tank, and at the same time, the camera takes pictures of the floating roof. After obtaining the collected image, the supplementary light is turned off;

[0011] S2: Transmit the collected image to the processor for processing, extract the color calibration gray card in the image, and adjust the color temperature, hue and brightness of the image based on the color calibration gray card;

[0012] S4: Scale down the adjusted image to a certain extent, and then convert the RGB color space in the image to the HSV color space;

[0013] S5: Identify the reference object on the floating roof and determine its position, and draw a rectangular frame and annotation on the image; When the floating roof is operating normally, the reference object does not change, and a safety signal is output and transmitted to the receiving terminal; When the reference object is identified as discolored or deformed, save the current image and transmit the image and the image of the floating roof tank body leakage to the receiving terminal;

[0014] S6: According to the known actual distance d between two reference objects, the viewing angle α of the lens, the pixel distance D between the two reference objects in the image, and the pixel length L of the long side of the picture, the distance h between the upper surface of the floating roof and the camera can be calculated, and then according to the distance h0 between the camera and the ground, the height H of the liquid level in the storage tank can be further obtained; Output the measurement result of the liquid level height H, and the specific formula is: Transmit the liquid level height H data to the receiving terminal;

[0015] S7: Compare the image pixels R1 and R2 of the diameters of the references at the two farthest ends of a diameter on the floating roof in the comparison screen. It is known that the references have the same actual diameter of r0. When the diameter pixels are the same, the floating roof is not tilted. When the image pixel R1 of the diameter of one side of the reference is greater than the diameter image pixel R2 of the other side, it indicates that the floating roof has tilted. The distance h1 between R1 and the camera is smaller than the distance h2 between R2 and the camera. Specifically, h1 / h2 = R2 / R1; the tilt angle is the detection result of the output floating roof tilt angle β. The specific formula is: Transmit the floating roof tilt angle β data to the receiving terminal;

[0016] S8: Display the leakage images, liquid level height measurement results, and floating roof tilt detection results received in S5, S6, and S7 on the display module after being processed and analyzed by the processor.

[0017] Furthermore, when the number of cameras in S1 is greater than 1, image stitching is required. The specific method is as follows: The image data collected by the cameras is transmitted to the processor. First, mark the spatial overlapping parts and their adjacent areas of the simultaneously captured images and divide them into a series of tiny sub-images. Extract the features contained in the corresponding sub-images as calibration anchor points; secondly, calculate the topological relationships of the anchor points with spatial correspondence and compare the mapping relationships of the topological networks. Calibrate the transformation matrices for translation, rotation, scale ratio transformation, affine, perspective, and barrel distortion respectively; obtain a series of spatially overlapping images of the topological network with exactly the same anchor points through matrix multiplication and connect them by means of anchor point positioning; finally, use the method of interpolation fitting to perform visual smoothing around the matching anchor points and at the seams, and output the stitched image.

[0018] Furthermore, the data, images, and results are transmitted to the receiving terminal through the 5G module.

[0019] Furthermore, the reference is specifically a substance or mechanical structure that will deform or change color when in contact with the oil product.

[0020] Furthermore, the substance is selected from test oil paste, sodium hydroxide, or discoloration test paper.

[0021] A vertical storage tank floating roof safety monitoring device based on machine vision includes an internal floating roof tank. There is a floating roof inside the internal floating roof tank. The upper end of the internal floating roof tank is a tank roof with an arc surface. The floating roof is composed of multiple box modules spliced together. There is a reference inside the box module of the floating roof. There is also a color calibration gray card on the floating roof. Several cameras and fill lights are arranged on the tank roof. The cameras and fill lights are connected to the processor, and the processor is connected to an external receiving end through the 5G module.

[0022] Furthermore, the external receiving end includes a mobile phone end and a computer end.

[0023] The following beneficial effects can be obtained through the above technical solutions:

[0024] During the operation of the oil tank for petroleum and its derivatives, information is captured by the camera, and there is no circuit operation inside the tank, ensuring the safety of the monitoring system itself; the monitored images are processed by machine vision methods, enabling round-the-clock and real-time monitoring of the operation status of the floating roof; a reference object is installed on each floating roof box module, enabling the system to not only monitor the overall operation status of the floating roof but also obtain the safety of each floating roof box module and understand the leakage situation in advance; the monitored data has multiple wireless output methods and can be remotely output to the receiving end, effectively reducing the labor intensity and working hours and facilitating the safe and efficient operation of the storage tank and floating roof.

[0025] Since there are a large number of floating boxes on the floating roof, one or two leaks have no impact. However, after the number of leaks accumulates, it will damage the operation safety. This system can detect the leakage situation before the floating roof tilts, greatly improving the operation safety of the fully liquid-tight internal floating roof. During shutdown maintenance, the leaking floating boxes can also be replaced according to the monitoring results, greatly reducing the maintenance cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Attached Figure 1 is a model diagram of the monitoring system;

[0027] Attached Figure 2 is a schematic diagram of the floating roof;

[0028] Attached Figure 3 is a hardware system architecture diagram;

[0029] Attached Figure 4 is the overall flowchart of the method. DETAILED DESCRIPTION OF THE INVENTION

[0030] The present invention will be further described below with reference to the accompanying drawings:

[0031] As Figure 4 shown, a method for safety monitoring of the floating roof of a vertical storage tank based on machine vision, characterized in that the method is as follows:

[0032] S1: The supplementary light provides illumination for the floating roof in the storage tank, and at the same time, the camera takes pictures of the floating roof. After obtaining the captured image, the supplementary light is turned off;

[0033] When the number of cameras is greater than 1, image stitching is required. The specific method is as follows: The image data collected by the cameras is transmitted to the processor. First, mark the spatial overlapping part and its adjacent areas of the images taken simultaneously, and divide them into a series of tiny sub-images. Extract the features contained in the corresponding sub-images as calibration anchor points. Secondly, calculate the topological relationship of the anchor points with spatial correspondence, and compare the mapping relationships of the topological networks. Calibrate the transformation matrices for translational, rotational, scale ratio transformation, affine, perspective, and barrel distortion respectively. Through matrix multiplication, a series of spatially overlapping images of the topological network with exactly the same anchor points are obtained, and they are connected by anchor point positioning. Finally, use the method of interpolation fitting to perform visual smoothing around the matching anchor points and at the seams, and output the stitched image;

[0034] S2: Transmit the collected images to the processor for processing, extract the color calibration gray card in the images, and adjust the color temperature, hue, and brightness of the images based on the color calibration gray card;

[0035] S4: Reduce the adjusted images to a certain extent, and then convert the RGB color space in the images to the HSV color space;

[0036] S5: Identify the reference objects on the floating roof and determine their positions, and draw rectangular frames and annotations on the images; When the floating roof is operating normally, the reference objects do not change, and a safety signal is output and transmitted to the receiving terminal; When the identified reference objects change color or deform, save the current image, and transmit the image and the image of the floating roof tank body with leakage to the receiving terminal;

[0037] S6: According to the known actual distance d between two reference objects, the viewing angle α of the lens, the pixel distance D between the two reference objects in the image, and the pixel length L of the long side of the picture, the distance h between the upper surface of the floating roof and the camera can be calculated. Then, based on the distance h0 between the camera and the ground, the height H of the liquid level in the storage tank can be further obtained; Output the measurement result of the liquid level height H. The specific formula is: Transmit the liquid level height H data to the receiving terminal;

[0038] S7: Compare the image pixels R1 and R2 of the diameters of the reference objects located at the two farthest ends of a diameter on the floating roof in the picture. It is known that the reference objects have the same actual diameter of r0. When the diameter pixels are the same, the floating roof has not tilted. When the image pixel R1 of the diameter of one side of the reference object is greater than the diameter image pixel R2 of the other side, it indicates that the floating roof has tilted. The distance h1 between R1 and the camera is smaller than the distance h2 between R2 and the camera. Specifically, h1 / h2 = R2 / R1; The tilt angle is the detection result of the tilt angle β of the floating roof. The specific formula is: Transmit the floating roof tilt angle β data to the receiving terminal;

[0039] S8: The leakage images, liquid level height measurement results, and floating roof tilt detection results received in S5, S6, and S7 are processed and analyzed by a processor and then displayed on the display module.

[0040] A safety monitoring device for the floating roof of a vertical storage tank based on machine vision includes an internal floating roof tank 1, the upper end of the internal floating roof tank is a tank top 2 with a circular arc surface, there is a floating roof 3 in the internal floating roof tank, the floating roof is composed of multiple box modules spliced together, and a reference object 4 is arranged inside the box module of the floating roof. A number of cameras 5 and supplementary lights 6 are arranged outside the tank, the multiple cameras and the supplementary lights are connected to a processor 7, the processor is connected to a 5G module 8, and data is transmitted to the equipment in the control room through 5G signals. Installation holes are provided at safe positions above the tank top or the tank body of the internal floating roof tank, and transparent explosion-proof glass is fixed in the installation holes to prevent oil and gas leakage. The cameras and the supplementary lights are installed outside the explosion-proof glass of the installation holes. The number of cameras and supplementary lights is designed according to the specific diameter, height of the storage tank and the floating roof, and the occlusion situation of internal objects. The minimum requirement is one top camera and one supplementary light. A color calibration gray card is installed on the upper surface of the floating roof to provide a benchmark for the camera shooting results. Small windows are opened on the upper surface of the box module of the floating roof, and transparent explosion-proof glass is fixed at the windows. The reference object is installed inside the explosion-proof glass on the inner side. The reference object is made of a substance or mechanism that will deform or change color when in contact with oil products, and does not require electric drive. For example, test oil paste, sodium hydroxide, or color-changing test paper is used. The specifications of the cameras are selected according to the specific diameter and height of the storage tank and the floating roof, and it is necessary to meet the requirement that the image effect of the reference object collected by the camera is clear. The specifications of the supplementary lights are selected according to the specific diameter and height of the storage tank and the floating roof, and it is necessary to meet the lighting intensity required by the camera. The processor development board can control the shooting of the camera and the lighting of the supplementary light to be synchronized, and can process the collected image data. The processor development board is connected to the 5G module and outputs the processed results through wireless signals. The 5G wireless signal can be received by a software terminal on a computer terminal 9 or a mobile terminal 10.

[0041] Example 1:

[0042] Figure 1-2As shown in the figure, a safety monitoring device for the floating roof of a vertical storage tank based on machine vision is introduced in detail. It includes an internal floating roof tank for storing gasoline (with a diameter of 10m and no internal obstruction). The upper end of the internal floating roof tank is a tank top with a circular arc surface. There is a floating roof in the internal floating roof tank, and the floating roof is composed of multiple box modules spliced together. There is a reference object inside the box module of the floating roof. The reference object is made of a substance that changes color when it comes into contact with the oil product and does not require electric drive. The initial color of the reference object is yellow and turns blue when it comes into contact with gasoline. One camera and one supplementary light are arranged outside the tank. The camera and the supplementary light are connected to a processor development board, with the model of Raspberry Pi 4B. The processor development board is connected to a 5G module, and the data is transmitted to the equipment in the control room through the 5G signal. The specification of the camera is selected as a model with a 4K resolution, and it is required to meet the requirement that the image of the reference object collected by the camera is clear. The specification of the supplementary light is selected as a 500w LED light, and the irradiation angle is diffused light, which can meet the required illumination intensity of the camera. The specific steps are as follows:

[0043] Step 1: The processor development board issues an instruction every 10 minutes to control the activation of one supplementary light and simultaneously commands one camera to take a picture.

[0044] Step 2: The image data collected by the camera is transmitted to the processor development board.

[0045] Step 3: Extract the color calibration gray card in the image and modify the color temperature, hue, and brightness of the image based on the color calibration gray card.

[0046] Step 4: Reduce the image to a certain extent, reduce the resolution from 4096*3112 to 1080*1920 to improve the operation speed. Convert the RGB color space in the image to the HSV color space.

[0047] Step 5: Identify the state of the reference object, find the position of the reference object, and draw a rectangular box and annotation on the image. When the floating roof is operating normally, the reference object does not change, and information indicating that the floating roof is operating normally is output and transmitted to the receiving terminal through the 5G module. When a reference object with a color change is identified, the current image is saved, and the image and the result of a leak in the floating roof box are transmitted to the receiving terminal through the 5G module.

[0048] Step 6: According to the known actual distance d between two reference objects, the viewing angle α of the lens, the pixel distance D between the two reference objects in the image, and the pixel length L of the long side of the picture, the distance h between the upper surface of the floating roof and the camera can be calculated, and then based on the distance h0 between the camera and the ground. The height H of the liquid level in the storage tank can be further obtained. Output the measurement result of the liquid level height H. The specific formula is:

[0049]

[0050] Step 7: Compare the image pixels R1 and R2 of the diameters of the references at the two farthest ends of a diameter on the floating roof in the picture. It is known that the references have the same actual diameter of r0. When the diameter pixels are the same, the floating roof is not tilted. When the image pixel R1 of the diameter of one side of the reference is greater than the diameter image pixel R2 of the other side, it indicates that the floating roof has tilted. The distance h1 between R1 and the camera is smaller than the distance h2 between R2 and the camera. Specifically, h1 / h2 = R2 / R1. The tilt angle is the detection result of the output floating roof tilt angle β. The specific formula is:

[0051] Step 8: Transmit the leakage detection result, the liquid level height measurement result, and the floating roof tilt detection result using the 5G module. The web server software on the processor development board can display the detection results on the web page, and can be remotely monitored and read through the browser software of the computer.

[0052] Embodiment 2:

[0053] If multiple cameras are used, S1: The fill light provides illumination for the floating roof in the storage tank, and multiple cameras take pictures of the floating roof. After obtaining the captured images, turn off the fill light. Image stitching is required. The specific method is as follows: The image data collected by the cameras is transmitted to the processor. First, mark the spatial overlapping parts and their adjacent areas of the images taken simultaneously, and divide them into a series of micro-sub-images. Extract the features contained in the corresponding sub-images as calibration anchor points. Secondly, calculate the topological relationship of the anchor points with spatial correspondence, and compare the mapping relationships of the topological networks. Respectively calibrate the transformation matrices of translation, rotation, scale ratio transformation, affine, perspective, and barrel distortion. Through matrix multiplication, a series of spatially overlapping images of the topological network with exactly the same anchor points are obtained, and the connection is achieved by anchor point positioning. Finally, use the method of interpolation fitting to perform visual smoothing around the matching anchor points and at the seams, and output the stitched image; S2: Transmit the captured image to the processor for processing, extract the color calibration gray card in the image, and adjust the color temperature, hue, and brightness of the image based on the color calibration gray card;

[0054] S4: Reduce the adjusted image to a certain extent, and then convert the RGB color space in the image to the HSV color space;

[0055] S5: Identify the references on the floating roof and determine their positions, and draw rectangular frames and annotations on the image. When the floating roof is operating normally, the references do not change, and a safety signal is output and transmitted to the receiving terminal. When a discolored or deformed reference is identified, save the current image, and transmit the image and the image of the leakage of the floating roof box body to the receiving terminal;

[0056] S6: Based on the known actual distance d between two references, the viewing angle α of the lens, the pixel distance D between the two references in the image, and the pixel length L of the long side of the picture, the distance h between the upper surface of the floating roof and the camera can be calculated. Then, based on the distance h0 between the camera and the ground, the height H of the liquid level in the storage tank can be further obtained. Output the measurement result of the liquid level height H. The specific formula is as follows: Transmit the liquid level height H data to the receiving terminal;

[0057] S7: Compare the image pixels R1 and R2 of the diameters of the references located at the two farthest ends of a diameter on the floating roof in the picture. It is known that the references have the same actual diameter of r0. When the diameter pixels are the same, the floating roof is not tilted. When the image pixel R1 of the diameter of one side of the reference is greater than the diameter image pixel R2 of the other side, it indicates that the floating roof has tilted. The distance h1 between R1 and the camera is smaller than the distance h2 between R2 and the camera. Specifically, h1 / h2 = R2 / R1. Output the detection result of the tilt angle β of the floating roof. The specific formula is as follows: Transmit the floating roof tilt angle β data to the receiving terminal;

[0058] S8: Display the leakage image, the liquid level height measurement result, and the floating roof tilt detection result received in S5, S6, and S7 on the display module after being processed and analyzed by the processor.

[0059] It should be noted that the reference is a comparison object. Its function is that when the oil leaks, after the oil contacts the reference, it will change color or deform, which is convenient for the camera to identify and judge. The reference of the present invention is not limited to the existing conventional substances, but can also be a mechanical structure or a sensor that can be triggered, as long as it can play an identification or development effect, etc., it can be used as a reference object.

[0060] The above are all the preferred embodiments of the present invention. For those of ordinary skill in the art of this technology, without departing from the principle of the present invention, any modifications to various equivalent forms of the present invention fall within the protection scope of the appended claims of this application.

Claims

1. A safety monitoring method for the floating roof of a vertical storage tank based on machine vision, characterized in that: The method is as follows: S1: The supplementary light provides illumination for the floating roof in the storage tank, and at the same time, the camera takes pictures of the floating roof. After obtaining the collected image, the supplementary light is turned off; S2: Transmit the collected image to the processor for processing, extract the color calibration gray card in the image, and adjust the color temperature, hue and brightness of the image based on the color calibration gray card; S4: Reduce the adjusted image to a certain extent, and then convert the RGB color space in the image to the HSV color space; S5: Identify the reference object on the floating roof and determine its position, and draw a rectangular frame and annotation on the image; when the floating roof is operating normally, the reference object does not change, output a safety signal and transmit it to the receiving terminal; when the reference object changes color or deforms, save the current image, and transmit the image and the image of the leakage of the floating roof tank body to the receiving terminal; S6: Based on the known actual distance d between two reference objects, the viewing angle α of the lens, the pixel distance D between the two reference objects in the image, and the pixel length L of the long side of the picture, the distance h between the upper surface of the floating roof and the camera can be calculated. Then, based on the distance h0 between the camera and the ground, the height H of the liquid level in the storage tank can be further obtained; output the measurement result of the liquid level height H, and the specific formula is: Transmit the liquid level height H data to the receiving terminal; S7: Compare the image pixels R1 and R2 of the diameters of the references at the two farthest ends of a diameter on the floating disc in the picture. It is known that the references have the same actual diameter of r0. When the diameter pixels are the same, the floating disc is not tilted. When the image pixel R1 of the diameter of one side of the reference is greater than the diameter image pixel R2 of the other side, it indicates that the floating disc has tilted. The distance h1 between R1 and the camera is smaller than the distance h2 between R2 and the camera. Specifically, h1 / h2 = R2 / R1; the tilt angle is the detection result of the output floating disc tilt angle β. The specific formula is: Transmit the floating disc tilt angle β data to the receiving terminal; S8: Display the leakage images, liquid level height measurement results, and floating roof tilt detection results received in S5, S6, and S7 on the display module after being processed and analyzed by the processor.

2. The method for safety monitoring of the floating roof of a vertical storage tank based on machine vision according to claim 1, wherein: When the number of cameras in S1 is greater than 1, image stitching is required. The specific method is as follows: The image data collected by the camera is transmitted to the processor. First, mark the spatial overlapping part and its adjacent areas of the images taken simultaneously, and divide them into a series of micro-sub-images. Extract the features contained in the corresponding sub-images as calibration anchor points; secondly, calculate the topological relationship of the anchor points with spatial correspondence, and compare the mapping relationship of the topological network to calibrate the transformation matrices of translation, rotation, scale ratio transformation, affine, perspective, and barrel distortion respectively; obtain a series of spatially overlapping images of the topological network with exactly the same anchor points through matrix multiplication, and connect them by anchor point positioning; finally, use the method of interpolation fitting to perform visual smoothing around the matching anchor points and at the seams, and output the stitched image.

3. A safety monitoring method for the floating roof of a vertical storage tank based on machine vision according to claim 1 or 2, characterized in that: Data, images, and results are transmitted to the receiving terminal through the 5G module.

4. A method for safety monitoring of the floating roof of a vertical storage tank based on machine vision according to claim 1 or 2, characterized in that: The reference object is specifically a substance or mechanical structure that will deform or change color when in contact with oil products.

5. A method for safety monitoring of the floating roof of a vertical storage tank based on machine vision according to claim 4, characterized in that: The substance uses oil testing paste or sodium hydroxide or litmus paper.

Images