A method for intelligently monitoring abnormal temperature of cryogenic storage tanks

By using visible light cameras and infrared cameras for image processing, the problem of high data quality and missed rate in temperature abnormality detection of low-temperature storage tanks is solved, which reduces computing resource consumption, improves environmental adaptability, and achieves efficient detection of temperature abnormalities .

CN118654766BActive Publication Date: 2024-12-03NINGBO OUYILE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411139619.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-12-03
Estimated Expiration
2044-08-20

AI Technical Summary

Technical Problem

The existing temperature anomaly detection technology has problems with data quality and integrity, high false alarm rate and omission rate, large computing resource consumption, poor generalization capability of model and insufficient dynamic environment adaptability, especially in low-temperature storage tank temperature monitoring. Effectively detect temperature abnormalities.

Method used

Image acquisition and processing are carried out using visible light cameras and infrared cameras. Through moving object shielding, image normalization, morphological black hat operation, contour extraction and feature screening, the temperature abnormality of low-temperature storage tanks is detected in real time, the omission rate is reduced, and the computing resources are reduced. Consumption, improve generalization ability and environmental adaptability.

Benefits of technology

It ensures the data quality and integrity of temperature anomaly detection, reduces the missed rate, reduces the consumption of computing resources, improves the generalization ability and environmental adaptability of the model, and can effectively detect temperature anomalies in various scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118654766B_ABST
    Figure CN118654766B_ABST
Patent Text Reader

Abstract

The present invention provides a method for intelligently monitoring temperature anomalies of cryogenic storage tanks, comprising: step S1, image acquisition: real-time acquisition of visible light images and infrared images; step S2, object shielding: performing moving object shielding processing on visible light images or infrared images; step S3, image normalization; step S4, black hat operation; step S5, segmentation operation: binarization processing is performed on low temperature anomaly points and high temperature anomaly points in the temperature anomaly area respectively, so as to extract small features generated by the black hat operation and include them in the binarized image; step S6, contour extraction; step S7, feature screening; step S8, temperature change alarm: obtaining the temperature change amount corresponding to each temperature change area, and outputting the position information and temperature change amount of each temperature change area to warn the operator. The beneficial effect is that the present invention can perform image detection processing for different day and night scenes, improve the generalization ability of the scene, and reduce the missed reporting rate of temperature anomaly points.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of infrared equipment fault detection, and in particular to a method for intelligently monitoring temperature anomalies of cryogenic storage tanks. Background Art

[0002] Temperature anomaly detection is a very important task in various application fields, especially in the fields of industry, environmental monitoring and health monitoring. Existing temperature anomaly detection technologies mainly rely on the following methods:

[0003] Statistical methods: Build a statistical model of temperature (such as mean, standard deviation, etc.) through historical data, and then detect anomalies based on these statistical indicators. For example, if a temperature reading deviates from the mean by more than a set standard deviation threshold, it may be considered an anomaly;

[0004] Machine learning and data mining techniques: Use machine learning algorithms (such as support vector machines, random forests, neural networks, etc.) to train models to identify abnormal patterns in temperature data. Data clustering and classification methods are also often used to detect anomalies by comparing data points with other data similarities or categories to determine whether they are abnormal;

[0005] Time series analysis: Time series analysis methods such as ARIMA (Autoregressive Integrated Moving Average) or Seasonally Decomposed Time Series (STL) are used to analyze the time trend and periodicity of temperature data to detect anomalies;

[0006] Internet of Things (IoT) technology: Real-time monitoring of temperature data using sensors and IoT devices that can collect data in real time and quickly detect anomalies using embedded algorithms.

[0007] The disadvantages of the above methods include:

[0008] (1) Data quality and integrity: If the input data is of poor quality or incomplete, the test results may be inaccurate. For example, sensor failure or errors in data transmission may affect the results.

[0009] (2) High false positive and false negative rates: Some methods may produce high false positive rates (incorrectly marking normal points as anomalies) or false negative rates (failing to detect true anomalies);

[0010] (3) Computing resources and real-time processing: Some complex machine learning models may require a lot of computing resources and may not be able to process large amounts of data in real time;

[0011] (4) Model generalization ability: Machine learning models may overfit to specific data sets and perform poorly on new or unseen data types;

[0012] (5) Dynamic environmental adaptability: In an environment where temperature data changes frequently, fixed statistical models or machine learning models may have difficulty adapting to environmental changes. Summary of the invention

[0013] The problem to be solved by the present invention is to provide a method for intelligently monitoring temperature anomalies in cryogenic storage tanks, which can ensure data integrity, reduce the underreporting rate of temperature anomalies, save computing resources, improve generalization ability, and enhance environmental adaptability.

[0014] In order to solve the above problems, the present invention provides a method for intelligently monitoring the abnormal temperature of a cryogenic storage tank. A bracket is fixedly installed on the side of the cryogenic storage tank to be detected in advance, and a camera barrel is fixedly installed on the bracket. A visible light camera and an infrared camera are arranged in the camera barrel, and the shooting ends of the visible light camera and the infrared camera are both facing the cryogenic storage tank to be detected. The method for intelligently monitoring the abnormal temperature of a cryogenic storage tank includes the following steps:

[0015] Step S1, image acquisition: acquiring in real time the visible light image output by the visible light camera and the infrared image output by the infrared camera;

[0016] Step S2, object shielding: determine whether the environment of the low-temperature storage tank to be detected is daytime:

[0017] If yes, performing moving object shielding processing on the visible light image, and converting the visible light image after the moving object shielding processing into a grayscale image for output;

[0018] If not, performing moving object shielding processing on the infrared image and outputting the image;

[0019] Step S3, image normalization: performing normalization processing on the infrared image or the grayscale image to obtain a normalized image and output it;

[0020] Step S4, black hat operation: performing a morphological black hat operation on the normalized image to obtain a black hat image containing a temperature abnormality region;

[0021] Step S5, segmentation operation: binarizing the low temperature abnormal points and the high temperature abnormal points in the temperature abnormal area in the black hat image according to a preset threshold value, so as to extract the small features generated by the black hat operation and include them in the binarized image;

[0022] Step S6, contour extraction: using the findContour function in the OpenCV visual algorithm library to extract the contours of each of the small features in the binary image;

[0023] Step S7, feature screening: traversing the small features after contour extraction according to pre-set screening parameters, screening out the small features that meet the screening parameters, and taking the contour corresponding to each small feature as the temperature change region;

[0024] Step S8, temperature change alarm: obtaining the temperature change amount corresponding to each of the temperature change areas, and outputting the position information and the temperature change amount of each of the temperature change areas to give an alarm prompt to the operator.

[0025] Preferably, in step S2, when the visible light image is subjected to moving object shielding processing or the infrared image is subjected to moving object shielding processing, a mixed Gaussian model in the OpenCV visual algorithm library is used to count the brightness value of each pixel on the visible light image or the infrared image in real time, and for the brightness value of each pixel, it is determined whether the brightness value exceeds the Gaussian distribution range:

[0026] If so, masking the pixel;

[0027] If not, the pixel is retained.

[0028] Preferably, in step S3, the infrared image or the grayscale image is a two-dimensional array, and the infrared image or the grayscale image is normalized by the following calculation formula:

[0029] ;

[0030] in,

[0031] Represents the normalized pixel value;

[0032] Indicates pixel value;

[0033] represents the minimum pixel value of the infrared image or the grayscale image;

[0034] Indicates the maximum pixel value of the infrared image or the grayscale image.

[0035] Preferably, in step S4, a morphological black hat operation is performed on the normalized image to remove the large background in the normalized image and retain each of the small features, each of which contains the temperature abnormality area.

[0036] Preferably, in step S5, set the pixel value in the image after the black hat operation as ir, set the maximum pixel value ir_max and the minimum pixel value ir_min of the image after the black hat operation, and the preset threshold th is between the maximum pixel value ir_max and the minimum pixel value ir_min. For the low-temperature abnormal points in the temperature abnormal area, the corresponding small features are extracted by binary processing ir < th, and for the high-temperature abnormal points in the temperature abnormal area, the corresponding small features are extracted by binary processing ir > th.

[0037] Preferably, the screening parameters include a size range, a roundness range, and a position range. The size range is 3 to 12 pixels, the roundness range is that the perimeter-area ratio is less than 50, and the position range is within a pre-defined area of interest. Then, in step S7, each of the small features that meet the requirements is screened according to the size range, the roundness range, and the position range.

[0038] Preferably, in step S8, region marking is performed and output in the visible light image or the infrared image according to the position information of each temperature change area.

[0039] The present invention has the following beneficial effects:

[0040] (1) The present invention uses a visible light camera and an infrared camera as image acquisition devices, which can cover a large spatial range, so that temperature abnormalities within the visible area of the camera can be detected, ensuring the quality and integrity of the data;

[0041] (2) The present invention makes improvements in response to the problems of high false alarm rate and missed detection rate in the prior art. Through hierarchical processing and screening of image normalization, black hat operation, segmentation operation, contour extraction, and feature screening, possible temperature abnormal areas are refined and the missed detection rate is reduced;

[0042] (3) The present invention does not require the use of complex machine learning models, and the data to be processed is only the visible light image output by the visible light camera and the infrared image output by the infrared camera. While ensuring functionality and real-time performance, the complexity and computational amount can be reduced;

[0043] (4) The visible light camera and the infrared camera, which are the image acquisition devices of the present invention, can adjust the installation position according to the size of different low-temperature storage tanks to be detected and the scenes. It has strong generalization ability for general industrial scenes and can realize the function of temperature abnormal detection in various occasions;

[0044] (5) In order to adapt to the temperature changes during the day and night, the present invention determines whether the environment of the low-temperature storage tank to be detected is day or night, and shields the moving objects in the corresponding image according to the judgment result. It can make corresponding corrections according to the scene changes and has strong environmental adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the positions between the low-temperature storage tank to be inspected and the visible light camera and the infrared camera of the present invention;

[0046] Figure 2 is a flow chart of the steps of the present invention;

[0047] Description of the accompanying drawings: 1. Low-temperature storage tank to be inspected; 2. Bracket; 3. Camera barrel; 4. Visible light camera; 5. Infrared camera. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0049] In a preferred embodiment of the present invention, based on the above problems existing in the prior art, a method for intelligently monitoring abnormal temperature of a cryogenic storage tank is provided. Figure 1 As shown, a bracket 2 is fixedly installed on the side of the low-temperature storage tank 1 to be detected in advance, and a camera barrel 3 is fixedly installed on the bracket 2. A visible light camera 4 and an infrared camera 5 are arranged in the camera barrel 3, and the shooting ends of the visible light camera 4 and the infrared camera 5 are both facing the low-temperature storage tank 1 to be detected. The method for intelligently monitoring the abnormal temperature of the low-temperature storage tank is as follows Figure 2 As shown, the following steps are included:

[0050] Step S1, image acquisition: acquiring in real time the visible light image output by the visible light camera 4 and the infrared image output by the infrared camera 5;

[0051] Step S2, object shielding: determine whether the environment of the low-temperature storage tank 1 to be detected is daytime:

[0052] If yes, then the visible light image is subjected to a moving object shielding process, and the visible light image after the moving object shielding process is converted into a grayscale image for output;

[0053] If not, the infrared image is processed for moving object shielding and output;

[0054] Step S3, image normalization: normalizing the infrared image or grayscale image to obtain a normalized image and output it;

[0055] Step S4, black hat operation: performing a morphological black hat operation on the normalized image to obtain a black hat image containing a temperature abnormality region;

[0056] Step S5, segmentation operation: binarizing the low temperature abnormal points and the high temperature abnormal points in the temperature abnormal area of ​​the black hat image according to a preset threshold value, so as to extract the small features generated by the black hat operation and include them in the binarized image;

[0057] Step S6, contour extraction: use the findContour function in the OpenCV visual algorithm library to extract the contours of each small feature in the binary image;

[0058] Step S7, feature screening: traverse the small features after contour extraction according to the pre-set screening parameters, screen out the small features that meet the screening parameters, and use the contours corresponding to the small features as the temperature change area;

[0059] Step S8, temperature change alarm: obtain the temperature change amount corresponding to each temperature change area, and output the position information and temperature change amount of each temperature change area to warn the operator.

[0060] Specifically, in this embodiment, considering that there may still be security and privacy issues in the prior art, especially when using IoT devices, data security and privacy protection become important issues. Therefore, the present invention can deploy a local file server for the low-temperature storage tank 1 to be detected, and store the data stream generated during the temperature anomaly detection process in the file server for backup. No network connection and IoT connection are required, thereby ensuring data security. Such backup operations can make it possible to trace back key events, reducing the impact of algorithm errors and software errors on the function of the low-temperature storage tank 1 to be detected.

[0061] Preferably, mathematical morphology is an image analysis discipline based on lattice theory and topology, and is the basic theory of mathematical morphology image processing. Its basic operations include: erosion and expansion, opening and closing operations, skeleton extraction, limit erosion, hit-miss transformation, morphological gradient, Top-hat transformation, particle analysis, watershed transformation, etc. The morphological black hat in the present invention is the image black hat operation, which is the result of the image closing operation minus the original image, to obtain a small hole inside the image, or a small black dot in the foreground color, which is expressed as:

[0062] Black hat operation(img) = closed operation image(img) - original image(img).

[0063] Preferably, the syntax format of the findContour function is:

[0064] image,contours,hierarchy=cv2.findcontours(image,mode, method);

[0065] In the formula, image represents the original image, an 8-bit single-channel image. All non-zero values ​​are processed as 1, and all zero values ​​remain unchanged. That is to say, the grayscale image will be automatically processed into a binary image. In actual operation, you can use threshold processing and other functions in advance to process the image to be searched for contours into a binary image as needed; mode represents the contour retrieval mode; method represents the contour approximation method.

[0066] Preferably, small features in a binary image can be described by global and local features of image features. Global features are statistical descriptions of the overall information of the image, such as color histograms, pyramid gradient oriented histograms (PHOG), GIST features, etc. These features can help identify the overall properties and structure of the image, while local features focus on local areas of the image. For small features in a binary image, the description of local features is particularly important. In a binary image, small features may include specific shapes, sizes, directions, and positions, etc. These features are of great significance in image processing and analysis.

[0067] Preferably, the present invention uses a visible light camera 4 and an infrared camera 5 as image acquisition devices. The visible light camera 4 adopts a passive receiving mode, does not need to provide an additional light source, and outputs an RGB format video stream; the infrared camera 5 works in a long-wave infrared band and adopts a passive receiving mode, does not need to provide an additional infrared light source, and outputs a grayscale video stream. The two cameras are installed in parallel at the front end of the outer shell of the low-temperature storage tank 1 to be inspected, and observe the same direction.

[0068] Specifically, in this embodiment, binarization is a common image processing technology, which achieves the black and white effect of the image by setting the grayscale value of the pixel on the image to 0 or 255. This processing method simplifies the data structure of the image, making the image more intuitive and convenient for subsequent analysis operations. In digital image processing, binary images play an important role because they can highlight the contours of the target of interest in the image while reducing the storage volume and computational complexity of the data. In this embodiment, the low-temperature anomaly points and high-temperature anomaly points in the temperature anomaly area are binarized separately to obtain the small features therein.

[0069] Preferably, when performing binarization processing, there are two main methods: a fixed threshold method and an adaptive threshold method. The fixed threshold method directly sets all pixels with grayscale values ​​greater than or equal to a certain threshold to 255 (white), and all pixels with grayscale values ​​less than the threshold to 0 (black). The adaptive threshold method dynamically selects a suitable threshold according to the characteristics of the image to obtain the best image segmentation effect.

[0070] Preferably, in practical applications, different threshold settings will produce different effects, so it is necessary to select the most appropriate method according to the specific image content and requirements.

[0071] Specifically, in this embodiment, in step S1, the visible light image and the infrared image are acquired in real time through the built-in computing unit using the opencv visual algorithm library. In step S6, a traversal method is used to find all contours, and then only the coordinates of the corners of the contours are given to reduce the amount of data.

[0072] In a preferred embodiment of the present invention, in step S2, when performing moving object shielding processing on a visible light image or performing moving object shielding processing on an infrared image, a mixed Gaussian model in the OpenCV visual algorithm library is used to count the brightness value of each pixel on the visible light image or the infrared image in real time, and for the brightness value of each pixel, it is determined whether the brightness value exceeds the Gaussian distribution range:

[0073] If so, mask the pixel;

[0074] If not, the pixel is retained.

[0075] Specifically, in this embodiment, considering that moving objects (such as people, vehicles, swaying trees, etc.) will interfere with any visual algorithm, they need to be shielded. Therefore, in this embodiment, the mixed Gaussian model (MOG2) in the OpenCV visual algorithm library is used and implemented after parameter optimization. MOG2 will count the brightness value of each pixel on each frame of infrared image or visible light image in real time and analyze and judge. When there is no moving object passing by, the brightness value of the pixel will be Gaussian distributed. When a moving object passes by, the brightness value of the pixel will exceed the Gaussian distribution range, and the pixel will be identified and shielded. In this way, all pixels passed by the moving object will be shielded, thereby achieving the effect of shielding the moving object. In this embodiment, the parameters of this method are optimized so that it can be applied to general industrial scenes. When the lighting conditions are good during the day, the visible light image information is more than the infrared image, and the visible light camera 4 is used to shield the infrared image for moving objects; when the lighting conditions are poor at night, the visible light camera 4 cannot capture the image, and the infrared camera 5 is used to shield the infrared image for moving objects to ensure the normal operation of the object shielding step.

[0076] In a preferred embodiment of the present invention, in step S3, the infrared image or grayscale image is a two-dimensional array, and the infrared image or grayscale image is normalized by the following calculation formula:

[0077] ;

[0078] in,

[0079] Represents the normalized pixel value;

[0080] Indicates pixel value;

[0081] Indicates the minimum pixel value of infrared image or grayscale image;

[0082] Indicates the maximum pixel value of an infrared image or grayscale image.

[0083] Specifically, in this embodiment, image normalization refers to the process of performing a series of standard processing transformations on an image to transform it into a fixed standard form. The standard image is called a normalized image. After the original image undergoes some processing or transformation, a variety of copy images can be obtained. These images can obtain standard images of the same form after image normalization processing with the same parameters. The main purpose of image normalization in this embodiment is to enhance the contrast of infrared images so that temperature abnormality areas can be easier to detect.

[0084] In a preferred embodiment of the present invention, in step S4, a morphological black hat operation is performed on the normalized image to remove the large background in the normalized image and retain various small features, each of which contains an abnormal temperature area.

[0085] Specifically, in this embodiment, morphological black hat operations are applied in step S4. Morphology, namely mathematical morphology, is a very important research direction in the image processing process. Morphology mainly extracts component information from the image. This component information is usually of great significance for expressing and depicting the shape of the image. It is usually the most essential shape feature used in image understanding. For example, when recognizing handwritten numbers, its skeleton information can be obtained through morphological operations. In specific recognition, only the skeleton needs to be operated. Morphological processing has very important applications in visual detection, text recognition, medical image processing, image compression coding and other fields.

[0086] Preferably, the morphological operations mainly include: erosion, dilation, opening operation, closing operation, morphological gradient operation, top-hat operation (hat operation), black-hat operation, etc. In this embodiment, the large background in the image is removed through the black-hat operation, leaving small feature information, which includes information on the temperature anomaly region.

[0087] Preferably, the black-hat operation is the result of subtracting the original image from the closing operation image of the image, obtaining small holes inside the image, or small black dots in the foreground color. Specifically: black-hat operation (img) = closing operation image (img) - original image (img).

[0088] In a preferred embodiment of the present invention, in step S5, the pixel value in the image after the black-hat operation is set as ir, the maximum pixel value ir_max and the minimum pixel value ir_min of the image after the black-hat operation are set, and the preset threshold th is between the maximum pixel value ir_max and the minimum pixel value ir_min. For the low-temperature anomaly points in the temperature anomaly region, the corresponding small features are extracted through binary processing ir < th, and for the high-temperature anomaly points in the temperature anomaly region, the corresponding small features are extracted through binary processing ir > th.

[0089] In a preferred embodiment of the present invention, the screening parameters include size range, roundness range, and position range. The size range is 3 to 12 pixels, the roundness range is that the perimeter-area ratio is less than 50, and the position range is within a pre-defined area of interest. Then, in step S7, the small features that meet the requirements are screened according to the size range, roundness range, and position range.

[0090] In a preferred embodiment of the present invention, in step S8, region marking is performed on the visible light image or infrared image according to the position information of each temperature change region and output.

[0091] Specifically, in this embodiment, in step S8, the temperature change amount is specifically manifested as the temperature difference between the start node and the end node within a time period, and the position information of the temperature change region is expressed using the pixel coordinates of the center point of each contour.

[0092] Preferably, in step S8, when performing region marking, the drawContours function in the opencv vision algorithm library is used to superimpose the found contour region on the visible light image or infrared image to achieve marking.

[0093] Although the present disclosure is disclosed as above, the protection scope of the present disclosure is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present disclosure, and these changes and modifications will all fall within the protection scope of the present invention.

Claims

1. A method for intelligently monitoring abnormal temperature of a cryogenic storage tank, characterized in that: A bracket (2) is fixedly installed in advance beside a low-temperature storage tank (1) to be inspected, and a camera barrel (3) is fixedly installed on the bracket (2), wherein a visible light camera (4) and an infrared camera (5) are arranged in the camera barrel (3), and the shooting ends of the visible light camera (4) and the infrared camera (5) are both facing the low-temperature storage tank (1) to be inspected. The method for intelligently monitoring the temperature abnormality of a low-temperature storage tank comprises the following steps: Step S1, image acquisition: acquiring in real time the visible light image output by the visible light camera (4) and the infrared image output by the infrared camera (5); Step S2, object shielding: determining whether the environment of the low-temperature storage tank (1) to be detected is daytime: If yes, performing moving object shielding processing on the visible light image, and converting the visible light image after the moving object shielding processing into a grayscale image for output; If not, performing moving object shielding processing on the infrared image and outputting the image; Step S3, image normalization: performing normalization processing on the infrared image or the grayscale image to obtain a normalized image and output it; Step S4, black hat operation: performing a morphological black hat operation on the normalized image to obtain a black hat image containing a temperature abnormality region; Step S5, segmentation operation: binarizing the low temperature abnormal points and the high temperature abnormal points in the temperature abnormal area in the black hat image according to a preset threshold value, so as to extract the small features generated by the black hat operation and include them in the binarized image; Step S6, contour extraction: using the findContour function in the OpenCV visual algorithm library to extract the contours of each of the small features in the binary image; Step S7, feature screening: traversing the small features after contour extraction according to pre-set screening parameters, screening out the small features that meet the screening parameters, and taking the contour corresponding to each small feature as the temperature change region; Step S8, temperature change alarm: obtaining the temperature change amount corresponding to each of the temperature change areas, and outputting the position information and the temperature change amount of each of the temperature change areas to give an alarm prompt to the operator.

2. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: In step S2, when the visible light image is subjected to moving object shielding processing or the infrared image is subjected to moving object shielding processing, the mixed Gaussian model in the OpenCV visual algorithm library is used to count the brightness value of each pixel on the visible light image or the infrared image in real time, and for the brightness value of each pixel, it is determined whether the brightness value exceeds the Gaussian distribution range: If so, masking the pixel; If not, the pixel is retained.

3. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: In step S3, the infrared image or the grayscale image is a two-dimensional array, and the infrared image or the grayscale image is normalized by the following calculation formula: ; in, Represents the normalized pixel value; Indicates pixel value; represents the minimum pixel value of the infrared image or the grayscale image; Indicates the maximum pixel value of the infrared image or the grayscale image.

4. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: In the step S4, a morphological black hat operation is performed on the normalized image to remove the large background in the normalized image and retain each of the small features, each of which contains the temperature abnormality area.

5. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: In step S5, the pixel point value in the image after the black hat operation is set as ir, the maximum pixel value ir_max and the minimum pixel value ir_min of the image after the black hat operation are set, and the preset threshold th is between the maximum pixel value ir_max and the minimum pixel value ir_min. For the low-temperature abnormal points in the temperature abnormal area, the corresponding small features are obtained by binary processing of ir < th, and for the high-temperature abnormal points in the temperature abnormal area, the corresponding small features are obtained by binary processing of ir > th.

6. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: The screening parameters include a size range, a roundness range, and a position range. The size range is 3 to 12 pixels, the roundness range is that the perimeter-area ratio is less than 50, and the position range is within a pre-defined area of interest. Then, in step S7, each of the small features that meet the requirements is screened according to the size range, the roundness range, and the position range.

7. The method for intelligently monitoring abnormal temperature of a cryogenic storage tank according to claim 1, characterized in that: In step S8, region marking is performed and output in the visible light image or the infrared image according to the position information of each temperature change area.