Method and device for identifying leakage of dangerous chemicals, and electronic device

Abstract

This disclosure provides embodiments of a method, apparatus, and electronic device for leak identification of hazardous chemicals. One specific implementation of the method includes: performing image correction on each thermal infrared image in a thermal infrared image sequence to generate a temperature field image, resulting in a temperature field image sequence; performing discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence; for each temperature field image group in the temperature field image group sequence, locating the leak area based on the temperature field image group to generate leak area information; filtering the temperature field image sequence based on the temperature field image group corresponding to the target leak area information to obtain a filtered temperature field image sequence; and identifying the leak trend based on the filtered temperature field image sequence to obtain leak trend information. This implementation achieves effective and timely leak identification during the dimethyl ether storage stage.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the field of chemical spill identification, particularly image recognition-based spill identification, and specifically to methods, apparatus, and electronic devices for identifying spills of hazardous chemicals. Background Technology

[0002] Dimethyl ether, also known as methyl ether or methoxymethane, is typically stored in large spherical tanks. Under standard conditions, it is a colorless, odorous, and flammable gas. When dimethyl ether leaks, it readily mixes with air to form an explosive mixture. Currently, leak detection during dimethyl ether storage typically relies on manual inspections or spherical tank pressure monitoring.

[0003] However, when using the above method, the following technical problems often arise:

[0004] (1) Manual inspection is difficult to effectively cover the surface of the spherical tank, making it difficult to identify dimethyl ether leaks in the blind spot in a timely manner;

[0005] (2) Dimethyl ether is sensitive to ambient temperature. Fluctuations in ambient temperature will directly cause changes in the pressure inside the spherical tank, which will mask the pressure changes in the early stage of leakage, thus making it impossible to identify dimethyl ether leakage in time through pressure changes. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure provide methods, apparatus, and electronic devices for identifying leaks of hazardous chemicals to address the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a method for identifying leaks of hazardous chemicals. The method includes: performing image correction on each thermal infrared image in a thermal infrared image sequence to generate a temperature field image, thereby obtaining a temperature field image sequence. The thermal infrared image sequence is acquired in real-time by a high-definition infrared camera facing the chemical storage tank, and the high-definition infrared camera covers both long-wave and mid-wave infrared bands. The temperature field image sequence is then discretely windowed to obtain a temperature field image group sequence, wherein a preset number of temperature field images are spaced between every two adjacent temperature field image groups in the temperature field image group sequence. For the temperature field image... For each temperature field image group in the sequence, the leak area is located based on the temperature field image group to generate leak area information, which includes: leak area coordinates and leak identifier; based on the temperature field image group corresponding to the target leak area information, the temperature field image sequence is filtered to obtain a filtered temperature field image sequence, where the target leak area information includes leak identifiers indicating leak areas where chemical leaks exist; based on the filtered temperature field image sequence, leak trend identification is performed to obtain leak trend information, which includes: a list of leak area coordinates, leak range trend information, and leak risk level.

[0009] Secondly, some embodiments of this disclosure provide a leak detection device for hazardous chemicals. The device includes: an image correction unit configured to perform image correction on each thermal infrared image in a thermal infrared image sequence to generate a temperature field image, thereby obtaining a temperature field image sequence, wherein the thermal infrared image sequence is acquired in real time by a high-definition infrared camera facing the chemical storage tank, the high-definition infrared camera covering the long-wave infrared band and the mid-wave infrared band; a discrete windowing unit configured to perform discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence, wherein a preset number of temperature field images are spaced between every two adjacent temperature field image groups in the temperature field image group sequence; and a leak area location unit configured to locate the leak area. Each temperature field image group in the temperature field image group sequence is used to locate the leak area to generate leak area information, which includes: leak area coordinates and leak identifier; an image filtering unit is configured to filter the temperature field image sequence according to the temperature field image group corresponding to the target leak area information to obtain a filtered temperature field image sequence, wherein the target leak area information includes leak identifiers that indicate the presence of chemical leaks; a leak trend identification unit is configured to identify leak trends according to the filtered temperature field image sequence to obtain leak trend information, wherein the leak trend information includes: a list of leak area coordinates, leak range trend information, and leak risk level.

[0010] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0011] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0012] The above embodiments of this disclosure have the following beneficial effects: Through the leakage identification method of some embodiments of this disclosure applied to hazardous chemicals, effective and timely leakage identification is achieved during the storage stage of dimethyl ether. Specifically, the reasons for the inability to effectively and timely identify leakage are: (1) Dimethyl ether is often stored in spherical tanks, and manual inspection is inefficient and makes it difficult to identify leakage locations in blind spots. (2) Dimethyl ether is sensitive to ambient temperature, and fluctuations in ambient temperature will directly cause changes in pressure inside the spherical tank, which will mask the pressure changes in the early stage of leakage, thus making it impossible to identify dimethyl ether leakage in time through pressure changes. Based on this, this disclosure considers that dimethyl ether will absorb a large amount of heat from the surroundings when leaking, resulting in obvious temperature changes at the leakage location. Therefore, this disclosure uses thermal infrared images as the basis for subsequent leakage identification. Specifically, firstly, image correction is performed on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, resulting in a temperature field image sequence. This thermal infrared image sequence is acquired in real-time by a high-definition infrared camera facing the chemical storage tank. This high-definition infrared camera covers both long-wave and mid-wave infrared bands. In practice, considering the image coverage, the cost of the infrared camera hardware, and the main characteristic absorption peaks of dimethyl ether, a high-definition infrared camera covering both long-wave and mid-wave infrared bands was chosen as the hardware basis for image acquisition. Image correction is then used to obtain the temperature field image representing the temperature value. Secondly, the temperature field image sequence is discretely windowed to obtain a temperature field image group sequence. A preset number of temperature field images are spaced between each two adjacent temperature field image groups in the temperature field image group sequence. In practice, to avoid missed detection of leaks, high-definition infrared cameras primarily operate in all-weather mode. However, leaks are low-probability events, resulting in a large number of invalid temperature field images. Considering that leaks are often continuous, discrete windowing effectively reduces the amount of data processing required for subsequent leak area localization. Next, for each temperature field image group in the aforementioned temperature field image sequence, leak area localization is performed based on the temperature field image group to generate leak area information, which includes leak area coordinates and a leak identifier. This achieves discrete leak area localization and identification. Further, based on the temperature field image group corresponding to the target leak area information, the aforementioned temperature field image sequence is filtered to obtain a filtered temperature field image sequence. The target leak area information includes leak identifiers representing leak areas where chemicals are leaking. Image filtering determines the main temperature field image intervals involved when a leak occurs. Finally, leak trend identification is performed based on the filtered temperature field image sequence to obtain leak trend information, which includes a list of leak area coordinates, leak range trend information, and leak risk level.This method enables effective and timely leakage identification during the storage of dimethyl ether. Attached Figure Description

[0013] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0014] Figure 1 This is a flowchart of some embodiments of a method for identifying leaks of hazardous chemicals according to the present disclosure;

[0015] Figure 2 This is a schematic diagram of the discrete windowing process;

[0016] Figure 3 This is a schematic diagram illustrating the process of determining windowing parameters;

[0017] Figure 4 This is a schematic diagram of the structure of some embodiments of a leak detection device for hazardous chemicals according to the present disclosure;

[0018] Figure 5 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0019] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0020] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0021] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0022] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0023] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0024] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of a method for identifying leaks of hazardous chemicals according to the present disclosure. This method for identifying leaks of hazardous chemicals includes the following steps:

[0026] Step 101: Perform image correction on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, thus obtaining a temperature field image sequence.

[0027] In some embodiments, the implementer (e.g., a computing device) of a leak detection method for hazardous chemicals can perform image correction on each thermal infrared image in a thermal infrared image sequence to generate a temperature field image, thus obtaining a temperature field image sequence.

[0028] The thermal infrared image sequence is acquired in real time by a high-definition infrared camera facing the chemical storage tank. The chemical storage tank is a storage device used to store hazardous chemicals. Specifically, the chemical storage tank can be a spherical tank used for storing dimethyl ether. The high-definition infrared camera can be a wide-angle infrared camera (field of view range of 60° ~ 120°). The resolution of the high-definition infrared camera is 1024 × 980. The aforementioned high-definition infrared camera covers the long-wave infrared band and the mid-wave infrared band. The long-wave infrared band can be 8μm ~ 14μm. The mid-wave infrared band can be 3μm ~ 5μm. The thermal infrared image represents the image formed by the thermal radiation energy emitted by the measured object received by the high-definition infrared camera. The temperature field image represents the image formed after the thermal radiation energy has undergone temperature conversion. The aforementioned thermal infrared image sequence is an ordered sequence. The aforementioned temperature field image sequence is an ordered sequence. There is a one-to-one correspondence between the thermal infrared images in the aforementioned thermal infrared image sequence and the temperature field images in the aforementioned temperature field image sequence.

[0029] In practice, considering the placement constraints of the spherical tank, the placement of the high-definition infrared cameras, and the need for comprehensive surface image acquisition, at least six high-definition infrared cameras can be used to capture images (thermal infrared images) of the tank surface. Specifically, taking into account engineering costs, six to eight high-definition infrared cameras are generally used to capture images of the tank surface.

[0030] In practice, firstly, the aforementioned executing entity can perform non-uniformity correction on the thermal infrared image. Next, radiometric conversion is performed on the digital quantization (DN) values ​​in the corrected thermal infrared image. Specifically, the DN values ​​can be converted into radiance (luminance) values ​​using the radiometric response function calibrated by the high-definition infrared camera. Further, temperature inversion is performed on the radiometrically converted thermal infrared image to obtain a temperature field image.

[0031] It should be noted that the aforementioned computing devices can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed within the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here. In particular, when the computing device is a terminal device, for example, it can be an edge computing device used to process data from multiple high-definition infrared cameras corresponding to a single chemical storage tank. Alternatively, it can be an edge computing device processing data from a single high-definition infrared camera.

[0032] In some optional implementations of certain embodiments, the execution entity performs image correction on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, including:

[0033] Step S1: Read the gain coefficient matrix and bias coefficient matrix corresponding to the above high-definition infrared camera.

[0034] The gain coefficient matrix stores the gain coefficients, and the bias coefficient matrix stores the bias coefficients. The dimensions of the gain coefficient matrix and the bias coefficient matrix are the same. The dimensions of the gain coefficient matrix are the same as the image resolution of the thermal infrared image. Each pixel (radiation) value in the thermal infrared image corresponds to a specific gain coefficient and bias coefficient.

[0035] In practice, after a high-definition infrared camera is calibrated, corresponding gain and bias coefficient matrices are obtained. To avoid redundant calculations of these two matrices, they are stored in the aforementioned computing device. That is, each high-definition infrared camera has its own gain and bias coefficient matrices. The computing device can retrieve the corresponding gain and bias coefficient matrices for each high-definition infrared camera based on its camera ID.

[0036] Step S2: Based on the above gain coefficient matrix and the above bias coefficient matrix, perform parallel pixel radiance correction on the above thermal infrared image to obtain the first corrected thermal infrared image.

[0037] The first corrected thermal infrared image is the thermal infrared image after pixel radiance value correction.

[0038] In practice, since each pixel (radiance) value in a thermal infrared image corresponds to a gain coefficient and a bias coefficient, for each pixel (radiance) value, the corrected pixel radiance value = pixel radiance value × the gain coefficient corresponding to the pixel radiance value + the bias coefficient corresponding to the pixel radiance value. Specifically, since it only involves pixel-by-pixel multiplication and addition operations, pixel radiance value correction can be performed on the pixel (radiance) values ​​in the thermal infrared image in parallel to obtain the first corrected thermal infrared image.

[0039] Step S3: Perform barrel distortion correction on the first corrected thermal infrared image to obtain the second corrected thermal infrared image.

[0040] The second corrected thermal infrared image is the first corrected thermal infrared image after distortion correction.

[0041] In practice, because high-definition infrared cameras are wide-angle infrared cameras (field of view range of 60° ~ 120°), the acquired thermal infrared images suffer from barrel distortion, necessitating barrel distortion correction. Specifically, since the image acquisition position and angle of the high-definition infrared camera are fixed, a pre-set position mapping table can be used for inverse position mapping. This table stores the pixel positions corresponding to the pixel (radio) values ​​before and after distortion correction. Finally, bilinear interpolation is used to obtain the pixel (radio) value at each pixel position in the second corrected thermal infrared image.

[0042] Step S4: Perform temperature inversion on the above-mentioned second corrected thermal infrared image to obtain the temperature field image corresponding to the above-mentioned thermal infrared image.

[0043] As an example, a lookup table between radiance (brightness) values ​​and temperature values ​​can be pre-built. This lookup table stores the mapping between radiance (brightness) values ​​and their corresponding temperature values ​​in the form of binary tuples. By looking up the two nearest temperature values ​​to the radiance value at each pixel location, the temperature value corresponding to that pixel location is determined through linear interpolation, thus obtaining the temperature field image. Because the lookup table between radiance (brightness) values ​​and temperature values ​​is pre-built, temperature inversion involves only one lookup and one interpolation, resulting in extremely high computational efficiency. This is particularly suitable for scenarios involving high-frequency image processing.

[0044] As another example, a polynomial fitting can be performed using the binary pairs stored in a lookup table between radiation (brightness) and temperature values ​​to obtain a polynomial fitting formula between them. The polynomial fitting process can employ the least squares method to solve for the coefficients in the polynomial fitting formula. By using the radiation (brightness) value as input to the polynomial fitting formula, the corresponding temperature value is obtained. Compared to the lookup table method, constructing a polynomial fitting formula quantifies the mapping relationship between radiation (brightness) and temperature values, requiring only a single polynomial solution during actual temperature inversion, resulting in higher computational efficiency.

[0045] Step 102: Discretize the temperature field image sequence to obtain a temperature field image group sequence.

[0046] In some embodiments, the execution entity described above can perform discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence.

[0047] In this temperature field image group sequence, a preset number of temperature field images are spaced between each two adjacent temperature field image groups. Each temperature field image group in the temperature field image group sequence contains the same number of temperature field images.

[0048] In practice, continuous image acquisition and leak identification of chemical storage tanks are required around the clock. However, chemical leaks are low-probability events, which means that most of the infrared images processed daily are invalid. In order to reduce the amount of data processing and the data processing pressure on computing devices, this disclosure reduces the number of temperature field images involved in the subsequent leak area location by using discrete windowing.

[0049] As an example, assuming the window length of the window is K (K≥3) and the sliding step size of the window is 2K, the temperature field image sequence can be discretely windowed by sliding the window to obtain a temperature field image group sequence. That is, every K (preset number) temperature field images, a temperature field image group containing K temperature field images is obtained, thus obtaining the temperature field image group sequence.

[0050] As an example, see Figure 2 The diagram illustrates the discrete windowing process, where the temperature field image sequence can consist of multiple temperature field images. Taking K = 3 as an example, the temperature field image group sequence can include: temperature field image group G1, temperature field image group G2, ..., temperature field image group Gn. Each temperature field image group includes 3 temperature field images. There is a 3-image interval between temperature field image groups G1 and G2.

[0051] In some optional implementations of certain embodiments, the execution entity performs discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence, including:

[0052] Step S1: Determine meteorological description information.

[0053] The meteorological description information characterizes the meteorological parameters of the area where the chemical storage tank is located. This meteorological description information includes: average wind speed, average temperature, average wind speed rate of change, and average temperature rate of change. The meteorological data acquisition period corresponding to the above meteorological description information is the same as the image period corresponding to the above thermal infrared image sequence. The average wind speed represents the average of multiple wind speed values ​​collected during the meteorological data acquisition period. The average temperature represents the average of multiple temperature values ​​collected during the meteorological data acquisition period. The average wind speed rate of change represents the average of the wind speed change rates corresponding to multiple wind speed values ​​collected during the meteorological data acquisition period. The average temperature rate of change represents the average of the temperature change rates corresponding to multiple temperature values ​​collected during the meteorological data acquisition period. The meteorological description information can be obtained through environmental temperature and wind speed sensors installed in the area where the chemical storage tank is located.

[0054] In practice, wind speed and temperature in the environment affect the expression of thermal radiation under different time periods or weather conditions. Compared with the method of using a fixed window length and sliding step size, this disclosure dynamically determines the subsequent windowing parameters by combining meteorological description information, thereby ensuring the rationality of the grouping of temperature field image groups and thus ensuring the robustness of subsequent leakage area location. In particular, since it is necessary to combine meteorological description information to determine the subsequent windowing parameters, it is necessary to ensure that the meteorological description information effectively represents the meteorological parameters within the image time period corresponding to the thermal infrared image sequence. Therefore, the meteorological data acquisition time period corresponding to the meteorological description information is the same as the image time period corresponding to the aforementioned thermal infrared image sequence. Specifically, the aforementioned execution entity can obtain the meteorological parameters collected by the environmental temperature acquisition sensor and the environmental wind speed acquisition sensor within the image time period corresponding to the infrared image sequence, thereby generating meteorological description information.

[0055] Step S2: Determine the window parameters based on the meteorological description information and the pre-constructed window parameter decision tree.

[0056] The windowing parameters mentioned above include: window length value and window spacing value.

[0057] Here, the windowing parameters represent the control parameters for the window length and window interval of the sliding window. The windowing parameter decision tree represents the pre-trained decision tree used to generate the windowing parameters corresponding to the meteorological description information. The window length value represents the window length of the sliding window. The window interval value represents the window interval between every two adjacent sliding windows.

[0058] In practice, this disclosure mainly uses wind speed and temperature as the basis for determining subsequent windowing parameters. There are four main scenarios: (1) low wind speed, high temperature; (2) low wind speed, low temperature; (3) high wind speed, high temperature; (4) high wind speed, low temperature. Specifically, firstly, since dimethyl ether absorbs a large amount of heat from its surroundings when leaking, the temperature difference is small when the ambient temperature is low, making it difficult to locate the leak area. Therefore, it is necessary to increase the window length and decrease the window interval to improve the robustness of locating the leak area. Secondly, when the ambient temperature is high, the temperature difference is large, and the boundary of the leak area is obvious. The window length and window interval can be reduced to ensure the robustness of locating the leak area while reducing the image processing load. Next, when the wind speed is low, the dimethyl ether concentration in the leak area is high, and the boundary of the leak area is obvious. The window length and window interval can be reduced. When the wind speed is high, the dimethyl ether in the leak area is dispersed, leading to a decrease in concentration. Therefore, it is necessary to increase the window length and decrease the window interval. This allows for the pre-training of a window parameter decision tree using historical meteorological descriptions and corresponding labeled windowing parameters as training samples. In particular, compared to manually setting rules, the constructed window parameter decision tree can automatically determine windowing parameters by combining meteorological descriptions. Furthermore, compared to using deep learning models, the model has lower training costs and higher practical value.

[0059] As an example, see Figure 3 The diagram illustrates the process of determining windowing parameters. The execution entity uses meteorological description information as input to the windowing parameter decision tree, and outputs windowing parameters corresponding to the meteorological description information. The windowing parameter decision tree consists of multiple decision nodes. During the pre-training phase, the decision conditions corresponding to each decision node are determined. The windowing parameter decision tree may include m leaf nodes, each leaf node corresponding to a windowing parameter. Figure 3 For example, the m leaf nodes of the window parameter decision tree can correspond to window parameters P1, P2, P3, ..., Pm-2, Pm-1, and Pm, respectively. The window parameter corresponding to each leaf node is a specific value for the window length and window interval.

[0060] Step S3: Construct a sliding window based on the window length value mentioned above.

[0061] The sliding window can be a moving window used to extract groups of temperature field images from a temperature field image sequence.

[0062] In practice, since the temperature field image sequence is a one-dimensional image sequence, a one-dimensional moving window can be constructed by combining the window length value, which serves as a sliding window.

[0063] Step S4: Use the sum of the above window length value and the above window interval value as the sliding step size.

[0064] Wherein, the sliding step size = window length value + window interval value.

[0065] Step S5: Based on the above sliding step size, control the above sliding window to perform sliding windowing on the above temperature field image sequence to obtain a temperature field image group sequence.

[0066] In practice, the position of the first temperature field image in the temperature field image sequence can be used as the starting position. Multiple temperature field images located within the sliding window can be used as a temperature field image group. The movement of the sliding window can be controlled by the sliding step size to obtain multiple temperature field image groups, which can be used as a temperature field image group sequence.

[0067] Step 103: For each temperature field image group in the temperature field image group sequence, locate the leakage area based on the temperature field image group to generate leakage area information.

[0068] In some alternative implementations of some embodiments, for each temperature field image group in the temperature field image group sequence, the aforementioned execution entity can locate the leakage area based on the temperature field image group to generate leakage area information.

[0069] The leakage area information is represented based on leakage areas identified from temperature field image groups. There is a correspondence between the leakage area information and the temperature field image groups. Assuming the temperature field image group sequence includes n temperature field image groups, each temperature field image group corresponds to one leakage area information, for a total of n leakage area information. The leakage area information includes: leakage area coordinates and a leakage identifier. The leakage area coordinates represent the image coordinates corresponding to the leakage area identified in the temperature field image. Specifically, the leakage area coordinates can be the coordinates of the corner point corresponding to the leakage area. The leakage identifier indicates whether a leakage area exists. For example, the leakage identifier can be represented by "1" or "0". When the leakage identifier is "1", it represents a leakage area identified in the temperature field image group. When the leakage identifier is "0", it represents a leakage area not identified in the temperature field image group.

[0070] In practice, a target detection model (such as the YOLO-V3 model) can be used to identify leakage areas in the temperature field images of the temperature field image group, thereby determining the leakage area information corresponding to the temperature field image group.

[0071] In some optional implementations of certain embodiments, the execution entity performs leakage region localization based on each temperature field image group in the temperature field image group sequence to generate leakage region information, including:

[0072] Step S1: Perform tank boundary identification on the first temperature field image in the above temperature field image group to obtain the tank boundary location.

[0073] The tank boundary location represents the boundary of the chemical storage tank in the temperature field image. Specifically, the tank boundary location can be represented by multiple boundary coordinates corresponding to the tank boundary. These boundary coordinates are image coordinates.

[0074] In practice, due to the uniform material of the outer surface of the storage tank, the tank has a clear thermal radiation boundary compared to the background area. Specifically, the tank boundary can be identified by using an edge detection algorithm based on the Canndy operator in the first temperature field image of the aforementioned temperature field image set, thus obtaining the location of the tank boundary.

[0075] Step S2: Determine the boundary deviation based on the above-mentioned tank boundary position and the tank calibration boundary position corresponding to the above-mentioned high-definition infrared camera.

[0076] The tank calibration boundary position represents the tank boundary corresponding to the chemical storage tank in the calibrated temperature field image. Specifically, the tank calibration boundary position can be represented by multiple boundary coordinates corresponding to the tank boundary. The boundary coordinates are image coordinates. The boundary deviation degree represents the degree of deviation of the tank boundary position from the tank calibration boundary position.

[0077] In practice, considering the potential for shooting angle shifts in high-definition infrared cameras during use (e.g., due to loose camera mounting), it's necessary to calculate whether the tank boundary in the temperature field image has shifted position, ensuring the reliability of the identified tank boundary location. Specifically, the distance between the tank's calibrated boundary position and the center point of the tank boundary position can be used as the boundary deviation. When the boundary deviation exceeds a preset deviation threshold, it indicates a significant shooting angle shift in the high-definition infrared camera. Even with further leak area localization, the change in the relative position between the high-definition infrared camera and the chemical tank can lead to inaccurate leak area localization (primarily due to coordinate deviation during the conversion from image coordinates to 3D world coordinates). Therefore, only when the boundary deviation is less than or equal to the preset deviation threshold can subsequent processing be effective. Furthermore, when the boundary deviation exceeds the preset deviation threshold, a camera calibration reminder can be directly sent to the maintenance terminal. This maintenance terminal can be a remote terminal specifically designed to monitor the operating status of the high-definition infrared camera.

[0078] Step S3: In response to the boundary deviation being less than or equal to a preset deviation threshold, construct a set of search anchor boxes based on the tank boundary position.

[0079] The search anchor box is a segmentation box used for image feature segmentation.

[0080] In practice, the area enclosed by the tank boundary can be used as the base area size. Multiple search anchor boxes can be constructed using preset scaling factors G (G≥3), resulting in a set of search anchor boxes. The scaling factor is used to scale the area enclosed by the tank boundary, and G can be 7. The scaling factor values ​​can be 1, 1.1, 1.2, 1.3, 1.4, 1.5, and 1.6. That is, the area enclosed by the tank boundary is proportionally enlarged using the scaling factor, and this is used as the search anchor box. The reason for this setting is that since leaks mainly occur on the surface of chemical storage tanks, from the perspective of the temperature field image, the leak area should appear around the area enclosed by the tank boundary, and the resulting thermal radiation changes should also appear around the area enclosed by the tank boundary. Based on this prior knowledge, by combining the tank boundary location to set the search anchor box set, compared to a full-image search method, the amount of invalid data processed can be effectively reduced.

[0081] Step S4: For each temperature field image in the above temperature field image group, perform the following positioning steps:

[0082] Step S41: Perform global feature extraction on the above temperature field image to obtain temperature field image features.

[0083] The temperature image features are the global image features corresponding to the temperature field image. The feature size of the temperature image features is the same as the image size of the temperature field image.

[0084] In practice, this disclosure uses an FPN (Feature Pyramid Networks) model to perform global feature extraction on a temperature field image, obtaining the temperature field image features. The FPN model includes 5 downsampling layers, 5 transposed convolutional layers, and 1 feature concatenation layer. Assume the temperature field image size is H×W. The first downsampling layer takes the temperature field image as input and outputs image features with a feature size of H / 2×W / 2. The second downsampling layer takes the output of the first downsampling layer as input and outputs image features with a feature size of H / 4×W / 4. The third downsampling layer takes the output of the second downsampling layer as input and outputs image features with a feature size of H / 8×W / 8. The fourth downsampling layer takes the output of the third downsampling layer as input and outputs image features with a feature size of H / 16×W / 16. The fifth downsampling layer takes the output of the fourth downsampling layer as input and outputs image features with a feature size of H / 32×W / 32. Each downsampling layer corresponds to a transposed convolutional layer, which reshapes the output of the corresponding downsampling layer to a feature size of H×W. The feature concatenation layer combines the five H×W image features output from the five transposed convolutional layers to obtain the temperature field image features. Specifically, the transposed convolutional layer corresponding to the first downsampling layer has an upsampling factor of 2, a transposed kernel size of 4×4, a stride of 2, and padding of 1.

[0085] The second downsampling layer corresponds to a transposed convolutional layer with an upsampling factor of 4, a transposed kernel size of 8×8, a stride of 4, and padding of 2. The third downsampling layer corresponds to a transposed convolutional layer with an upsampling factor of 8, a transposed kernel size of 16×16, a stride of 8, and padding of 4. The fourth downsampling layer corresponds to a transposed convolutional layer with an upsampling factor of 16, a transposed kernel size of 32×32, a stride of 16, and padding of 8. The fifth downsampling layer corresponds to a transposed convolutional layer with an upsampling factor of 32, a transposed kernel size of 64×64, a stride of 32, and padding of 16. Each transposed convolutional layer operates independently, ensuring that feature upsampling is performed independently on image features extracted from different receptive fields.

[0086] Step S42: Based on the above search anchor box set, perform feature segmentation on the above temperature field image features to obtain a local temperature field image feature set.

[0087] The local temperature field image features are the local image features located within the search anchor box. The search anchor box is a segmentation box used for image feature segmentation.

[0088] In practice, the aforementioned executing entity can segment local image features located within the search anchor box in the temperature field image features and use them as local temperature field image features.

[0089] Step S43: Perform parallel leakage area localization based on the above local temperature field image feature set to obtain initial leakage area information.

[0090] Each local temperature field image feature corresponds to a leak location head. The leak location head takes the corresponding local temperature field image feature as input and outputs the corresponding leak region coordinates and leak identifier. The number of leak location heads is consistent with the number of scaling factors. Each leak location head consists of a downsampling layer, a location regressor, and a leak identifier classifier. The downsampling layer in the leak location head is used to shape the local temperature field image features, using the shaped image features as input to the location regressor and leak representation classifier. The location regressor outputs the leak region coordinates. The leak identifier classifier (binary classifier) ​​outputs the leak identifier. Multiple leak location heads corresponding to the local temperature field image feature set are set up in parallel, thus generating multiple sets of leak region coordinates and multiple sets of leak identifiers. Each leak region coordinate has a corresponding confidence level, and each leak representation has a corresponding confidence level. By averaging and weighting the confidence levels, the leak region coordinates and leak identifier with the highest confidence are selected from the multiple sets of leak region coordinates and multiple sets of leak identifiers as the initial leak region information.

[0091] Step S5: Based on the obtained initial leakage area information set, generate leakage area information corresponding to the above temperature field image set.

[0092] In practice, when a predetermined proportion of the initial leak area information set contains leak identifiers representing leak areas identified in the temperature image group, the leak identifiers representing the leak areas identified in the temperature image group are used as the leak identifiers included in the leak area information corresponding to the aforementioned temperature field image group. Furthermore, the coordinates of the leak areas included in the initial leak area information set are used as the coordinates of the leak areas included in the leak area information.

[0093] In practice, single temperature field image recognition may result in misidentification or missed identification due to environmental factors. Therefore, by combining the initial leakage area information set corresponding to the temperature field image group, the probability of misidentification and missed identification can be reduced, thereby forming leakage area information that characterizes whether the temperature field image group is a leakage area that has been identified.

[0094] Step 104: Based on the temperature field image group corresponding to the target leakage area information, perform image filtering on the temperature field image sequence to obtain the filtered temperature field image sequence.

[0095] In some embodiments, the aforementioned execution entity can perform image filtering on the temperature field image sequence based on the temperature field image group corresponding to the target leakage area information to obtain a filtered temperature field image sequence.

[0096] The target leak area information includes leak identification markings indicating the leak area where a chemical leak has occurred.

[0097] In practice, chemical leaks often occur as continuous leaks, resulting in multiple leak area information points, including leak markers, indicating the presence of a chemical leak. Therefore, by determining the relative position of the temperature field image group corresponding to the target leak area within the temperature field image sequence, most of the temperature field images related to the chemical leak can be identified, serving as the filtered temperature field image sequence.

[0098] As an example, see further. Figure 2 Assuming that the leakage area information corresponding to temperature field image group G2 includes leakage markers indicating a chemical leak, and the leakage area information corresponding to temperature field image group Gn includes leakage markers indicating a chemical leak, then the image position of the first temperature field image in temperature field image group G2 can be taken as the starting position and the image position of the last temperature field image in temperature field image group Gn can be taken as the ending position. The temperature field images in the temperature field image sequence located between the starting position and the ending position can be used as the filtered temperature field image sequence.

[0099] In some optional implementations of certain embodiments, the execution entity performs image filtering on the temperature field image sequence based on the temperature field image group corresponding to the target leakage area information, to obtain a filtered temperature field image sequence, including:

[0100] Step S1: Determine the starting and ending screening boundary positions based on the temperature field image group corresponding to the target leakage area information.

[0101] The starting and ending screening boundary positions represent the starting and ending positions used to segment the temperature field image sequence.

[0102] In practice, since there may be multiple leak area information, including leak markers indicating chemical leaks, there may be multiple temperature field image groups corresponding to the target leak area information. Therefore, the minimum image position corresponding to the temperature field image in the multiple temperature field image groups corresponding to the target leak area information can be used as the starting screening boundary position, and the maximum image position corresponding to the temperature field image can be used as the ending screening boundary position.

[0103] Step S2: Based on the above-mentioned starting screening boundary position and the above-mentioned ending screening boundary position, the above-mentioned temperature field image sequence is segmented to obtain the filtered temperature field image sequence.

[0104] In practice, the aforementioned executing entity can use multiple temperature field images in the temperature field image sequence that are located before the starting screening boundary and the aforementioned ending screening boundary as the filtered temperature field image sequence.

[0105] Step 105: Based on the filtered temperature field image sequence, leakage trend identification is performed to obtain leakage trend information.

[0106] In some embodiments, the aforementioned execution entity can identify leakage trends based on the filtered temperature field image sequence to obtain leakage trend information.

[0107] The leakage trend information characterizes the leakage trend of the chemical leak. This information includes: a list of leakage area coordinates, leakage extent trend information, and leakage risk level. The leakage area coordinates in the list represent the changing boundaries of the leakage area over time. The leakage extent trend information represents the regional change trend of the leakage area. The leakage risk level characterizes the hazard level of the chemical leak.

[0108] In practice, firstly, a leak location can be identified temporally using a model such as YOLO, taking a filtered temperature field image sequence as input, to obtain a list of leak area coordinates. Then, based on this list, the leak area corresponding to each time granularity is determined, and the size change between adjacent leak areas is calculated to obtain leak range trend information. Finally, a level mapping is performed based on the size of the leak area to obtain the leak risk level.

[0109] In some optional implementations of certain embodiments, the execution entity performs leakage trend identification based on the filtered temperature field image sequence to obtain leakage trend information, including:

[0110] Step S1: For each filtered temperature field image in the above filtered temperature field image sequence, locate the leakage area of ​​the above filtered temperature field image to obtain the leakage area coordinates corresponding to the above filtered temperature field image in the leakage area coordinate list.

[0111] In practice, firstly, when the filtered temperature field image is within a temperature field image group, leakage area localization is no longer performed. Instead, the leakage area coordinates included in its corresponding leakage area information are directly used as the leakage area coordinates in the leakage area coordinate list corresponding to the filtered temperature field image. When the filtered temperature field image is not within a temperature field image group, it indicates that the filtered temperature field image is between two temperature field image groups. Therefore, steps S41 to S43 in step 103 can be directly reused to determine the leakage area coordinates corresponding to the filtered temperature field image (the leakage area coordinates with the highest confidence included in the initial leakage area information). In particular, when the filtered temperature field image is not within a temperature field image group, the set of search anchor boxes used for the filtered temperature field image is the set of search anchor boxes generated by the last temperature field image in the previous temperature field image group, thereby avoiding the regeneration of search anchor boxes and improving search efficiency.

[0112] Step S2: Determine the sequence of leak range areas based on the list of leak area coordinates.

[0113] The leakage range region represents the leakage area defined by the coordinates of the leakage regions. There is a one-to-one correspondence between the leakage region coordinates in the leakage region coordinate list and the leakage range regions in the leakage range region sequence.

[0114] In practice, since the coordinates of the leaked area can be the coordinates of the corner points corresponding to the leaked area, for each leaked area coordinate in the list of leaked area coordinates, the closed area enclosed by the leaked area coordinates can be taken as the leak range area, thus obtaining the sequence of leak range areas.

[0115] Step S3: Based on the above leakage range area sequence, determine the leakage range trend information included in the leakage trend information.

[0116] In practice, the aforementioned executing entity can calculate the ratio of the area size of every two adjacent leakage range areas (the area size of the latter leakage range area / the area size of the former leakage range area) to obtain multiple time-series area size ratios, which can be used as leakage range trend information.

[0117] Step S4: Map the risk level based on the above leakage range trend information to obtain the leakage risk level included in the leakage trend information.

[0118] In practice, the leakage range trend information is actually a one-dimensional sequence of region size ratios. Therefore, a pre-trained multi-classifier can be used to classify the leakage risk level using the leakage range trend information as input. The multi-classifier consists of multiple downsampling layers and one softmax activation function.

[0119] In some optional implementations of some embodiments, the above method further includes:

[0120] Step S1: Perform coordinate transformation on the list of leak area coordinates included in the above leak trend information to obtain the three-dimensional leak area coordinates.

[0121] The three-dimensional leak area coordinates represent the three-dimensional coordinates of the corner points corresponding to the leak area in the geodetic coordinate system. In practice, the high-definition infrared camera can be pre-calibrated to determine its rotation matrix and translation vector. Then, the leak area coordinates in the leak area coordinate list are mapped to three-dimensional coordinates using the rotation matrix and translation vector, serving as the three-dimensional leak area coordinates. In particular, considering that the leak area usually expands when a chemical leak occurs, only the last leak area coordinate in the leak area coordinate list needs to be transformed to obtain the three-dimensional leak area coordinates.

[0122] Step S2: Based on the dimethyl ether gas sensor topology diagram, determine the dimethyl ether gas sensor corresponding to the coordinates of the three-dimensional leakage area mentioned above, and use it as the target dimethyl ether gas sensor.

[0123] The aforementioned dimethyl ether gas sensor topology diagram represents the location topology of dimethyl ether gas sensors installed around the chemical storage tank. The dimethyl ether gas sensor is used to detect the presence of dimethyl ether gas.

[0124] In practice, dimethyl ether (DME) gas sensors are relatively expensive, and due to the large size of the spherical tank, the traditional method of using multiple DME gas sensors in a discrete array cannot cover the entire surface of the tank without any blind spots. Therefore, this solution only places a small number of DME gas sensors at specific locations around the spherical tank to reduce equipment costs. Each DME gas sensor can be pre-set with its corresponding signal acquisition range and location, thus forming a DME gas sensor topology. By combining the coordinates of the three-dimensional leakage area with the corresponding signal acquisition range of the DME gas sensor, the target DME gas sensor is determined.

[0125] Step S3: In response to the leakage risk level being classified as the first risk level in the above leakage trend information, read the historical gas concentration signal collected by the target dimethyl ether gas sensor.

[0126] The signal acquisition period corresponding to the aforementioned historical gas concentration signal is the same as the image period corresponding to the aforementioned thermal infrared image sequence. The historical gas concentration signal represents the dimethyl ether gas concentration signal acquired by the target dimethyl ether gas sensor within the corresponding signal acquisition range.

[0127] In practice, since this solution mainly relies on thermal infrared images for leak identification, in order to avoid the problem of insufficient identification accuracy of a single mode under special circumstances, a dimethyl ether gas sensor is further introduced to collect gas concentration signals for risk level updates, thereby improving the accuracy of leak identification from a multimodal perspective.

[0128] Step S4: In response to the leakage risk level being the second risk level included in the leakage trend information, increase the signal acquisition frequency of the target dimethyl ether gas sensor and read the real-time gas concentration signal acquired by the target dimethyl ether gas sensor.

[0129] The first risk level is lower than the second risk level. When the risk level identified by combining thermal infrared images is low, the risk level is updated only based on historical gas concentration signals for the same period. When the risk level identified by combining thermal infrared images is high, the risk level is updated promptly based on real-time gas concentration signals.

[0130] In practice, the aforementioned executing entity can send control commands to the target dimethyl ether gas sensor to increase the signal acquisition frequency, and obtain the real-time gas concentration signal returned by the target dimethyl ether gas sensor after the signal acquisition frequency is increased.

[0131] Step S5: In response to the presence of signal anomalies in the aforementioned historical gas concentration signal or the aforementioned real-time gas concentration signal, update the leakage risk level included in the aforementioned leakage trend information.

[0132] In practice, a gas concentration threshold can be preset. By comparing the historical gas concentration signal or the real-time gas concentration signal with the gas concentration threshold, it can be determined whether there is a signal anomaly. When a signal anomaly is found, that is, when there is a signal value in the historical gas concentration signal or the real-time gas concentration signal that is greater than the gas concentration threshold, the leakage risk level included in the leakage trend information is increased.

[0133] The above embodiments of this disclosure have the following beneficial effects: Through the leakage identification method of some embodiments of this disclosure applied to hazardous chemicals, effective and timely leakage identification is achieved during the storage stage of dimethyl ether. Specifically, the reasons for the inability to effectively and timely identify leakage are: (1) Dimethyl ether is often stored in spherical tanks, and manual inspection is inefficient and makes it difficult to identify leakage locations in blind spots. (2) Dimethyl ether is sensitive to ambient temperature, and fluctuations in ambient temperature will directly cause changes in pressure inside the spherical tank, which will mask the pressure changes in the early stage of leakage, thus making it impossible to identify dimethyl ether leakage in time through pressure changes. Based on this, this disclosure considers that dimethyl ether will absorb a large amount of heat from the surroundings when leaking, resulting in obvious temperature changes at the leakage location. Therefore, this disclosure uses thermal infrared images as the basis for subsequent leakage identification. Specifically, firstly, image correction is performed on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, resulting in a temperature field image sequence. This thermal infrared image sequence is acquired in real-time by a high-definition infrared camera facing the chemical storage tank. This high-definition infrared camera covers both long-wave and mid-wave infrared bands. In practice, considering the image coverage, the cost of the infrared camera hardware, and the main characteristic absorption peaks of dimethyl ether, a high-definition infrared camera covering both long-wave and mid-wave infrared bands was chosen as the hardware basis for image acquisition. Image correction is then used to obtain the temperature field image representing the temperature value. Secondly, the temperature field image sequence is discretely windowed to obtain a temperature field image group sequence. A preset number of temperature field images are spaced between each two adjacent temperature field image groups in the temperature field image group sequence. In practice, to avoid missed detection of leaks, high-definition infrared cameras primarily operate in all-weather mode. However, leaks are low-probability events, resulting in a large number of invalid temperature field images. Considering that leaks are often continuous, discrete windowing effectively reduces the amount of data processing required for subsequent leak area localization. Next, for each temperature field image group in the aforementioned temperature field image sequence, leak area localization is performed based on the temperature field image group to generate leak area information, which includes leak area coordinates and a leak identifier. This achieves discrete leak area localization and identification. Further, based on the temperature field image group corresponding to the target leak area information, the aforementioned temperature field image sequence is filtered to obtain a filtered temperature field image sequence. The target leak area information includes leak identifiers representing leak areas where chemicals are leaking. Image filtering determines the main temperature field image intervals involved when a leak occurs. Finally, leak trend identification is performed based on the filtered temperature field image sequence to obtain leak trend information, which includes a list of leak area coordinates, leak range trend information, and leak risk level.This method enables effective and timely leakage identification during the storage of dimethyl ether.

[0134] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a leak detection device for hazardous chemicals, which are similar to... Figure 1 Corresponding to the method embodiments shown, this leak detection device for hazardous chemicals can be specifically applied to various electronic devices.

[0135] like Figure 4 As shown, some embodiments of a leak detection device 400 for hazardous chemicals include: an image correction unit 401, a discrete windowing unit 402, a leak area location unit 403, an image filtering unit 404, and a leak trend identification unit 405. The image correction unit 401 is configured to perform image correction on each thermal infrared image in a thermal infrared image sequence to generate a temperature field image, resulting in a temperature field image sequence. This thermal infrared image sequence is acquired in real-time by a high-definition infrared camera facing the chemical storage tank, and the high-definition infrared camera covers both long-wave and mid-wave infrared bands. The discrete windowing unit 402 is configured to perform discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence. The temperature field image group sequence has a preset number of images spaced apart between each two adjacent temperature field image groups. Temperature field images; a leak area location unit 403 is configured to locate the leak area for each temperature field image group in the temperature field image group sequence to generate leak area information, wherein the leak area information includes: leak area coordinates and leak identifier; an image filtering unit 404 is configured to filter the temperature field image sequence according to the temperature field image group corresponding to the target leak area information to obtain a filtered temperature field image sequence, wherein the target leak area information includes leak identifiers indicating a leak area where a chemical leak exists; a leak trend identification unit 405 is configured to identify the leak trend according to the filtered temperature field image sequence to obtain leak trend information, wherein the leak trend information includes: a list of leak area coordinates, leak range trend information, and leak risk level.

[0136] It is understood that the units described in the leak detection device 400 for hazardous chemicals are consistent with the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the leak detection device 400 for hazardous chemicals and the units contained therein, and will not be repeated here.

[0137] The following is for reference. Figure 5It shows a schematic diagram of the structure of an electronic device (e.g., a computing device) 500 suitable for implementing some embodiments of the present disclosure. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0138] like Figure 5 As shown, the electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory 502 or a program loaded from a storage device 508 into a random access memory 503. The random access memory 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, the read-only memory 502, and the random access memory 503 are interconnected via a bus 504. An input / output interface 505 is also connected to the bus 504.

[0139] Typically, the following devices can be connected to the input / output interface 505: input devices 506 including, for example, a touchscreen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 508 including, for example, magnetic tape, hard disk, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 5 Each box shown can represent a device or multiple devices as needed.

[0140] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a read-only memory 502. When the computer program is executed by the processing device 501, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0141] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0142] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0143] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: perform image correction on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, thereby obtaining a temperature field image sequence, wherein the thermal infrared image sequence is acquired in real time by a high-definition infrared camera facing the chemical storage tank, the high-definition infrared camera covering the long-wave infrared band and the mid-wave infrared band; perform discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence, wherein a preset number of temperature field images are spaced between every two adjacent temperature field image groups in the temperature field image group sequence; for the aforementioned Each temperature field image group in the temperature field image group sequence is used to locate the leak area to generate leak area information, which includes leak area coordinates and leak identifiers. Based on the temperature field image group corresponding to the target leak area information, the temperature field image sequence is filtered to obtain a filtered temperature field image sequence. The target leak area information includes leak identifiers indicating leak areas where chemicals are leaking. Based on the filtered temperature field image sequence, leak trend identification is performed to obtain leak trend information, which includes a list of leak area coordinates, leak range trend information, and leak risk level.

[0144] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and Python, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0145] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0146] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0147] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for identifying leaks of hazardous chemicals, characterized in that, include: Each thermal infrared image in the thermal infrared image sequence is image corrected to generate a temperature field image, resulting in a temperature field image sequence. The thermal infrared image sequence is acquired in real time by a high-definition infrared camera facing the chemical storage tank, and the high-definition infrared camera covers the long-wave infrared band and the mid-wave infrared band. Discrete windowing is performed on the temperature field image sequence to obtain a temperature field image group sequence, wherein a preset number of temperature field images are spaced between every two adjacent temperature field image groups in the temperature field image group sequence; For each temperature field image group in the temperature field image group sequence, the leakage area is located based on the temperature field image group to generate leakage area information, wherein the leakage area information includes: leakage area coordinates and leakage identifier; Based on the temperature field image group corresponding to the target leakage area information, the temperature field image sequence is filtered to obtain a filtered temperature field image sequence, wherein the target leakage area information includes leakage identifiers that characterize leakage areas where chemical leakage exists; Based on the filtered temperature field image sequence, leakage trend identification is performed to obtain leakage trend information, which includes: a list of leakage area coordinates, leakage range trend information, and leakage risk level.

2. The method for identifying leaks of hazardous chemicals according to claim 1, characterized in that, The method further includes: The list of leakage area coordinates included in the leakage trend information is transformed to obtain three-dimensional leakage area coordinates; Based on the dimethyl ether gas sensor topology map, the dimethyl ether gas sensor corresponding to the coordinates of the three-dimensional leakage area is determined as the target dimethyl ether gas sensor. The dimethyl ether gas sensor topology map represents the position topology of the dimethyl ether gas sensors installed around the chemical storage tank. In response to the leakage trend information including a leakage risk level of the first risk level, the historical gas concentration signal collected by the target dimethyl ether gas sensor is read, wherein the signal acquisition period corresponding to the historical gas concentration signal is the same as the image period corresponding to the thermal infrared image sequence; In response to the leakage trend information indicating a leakage risk level of Level 2, the signal acquisition frequency of the target dimethyl ether gas sensor is increased, and the real-time gas concentration signal acquired by the target dimethyl ether gas sensor is read. In response to an anomaly in the historical gas concentration signal or the real-time gas concentration signal, the leakage risk level included in the leakage trend information is updated.

3. The method for identifying leaks of hazardous chemicals according to claim 2, characterized in that, The step of performing image correction on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image includes: Read the gain coefficient matrix and bias coefficient matrix corresponding to the high-definition infrared camera, wherein the matrix dimension of the gain coefficient matrix is ​​the same as that of the bias coefficient matrix, and the matrix dimension of the gain coefficient matrix is ​​the same as that of the thermal infrared image. Based on the gain coefficient matrix and the bias coefficient matrix, the thermal infrared image is subjected to parallel pixel radiance correction to obtain a first corrected thermal infrared image. The first corrected thermal infrared image is subjected to barrel distortion correction to obtain a second corrected thermal infrared image; Temperature inversion is performed on the second corrected thermal infrared image to obtain the temperature field image corresponding to the thermal infrared image.

4. The method for identifying leaks of hazardous chemicals according to claim 3, characterized in that, The step of discretizing and windowing the temperature field image sequence to obtain a temperature field image group sequence includes: Determine meteorological description information, wherein the meteorological description information includes: average wind speed, average temperature, average wind speed change rate, and average temperature change rate, and the meteorological data collection period corresponding to the meteorological description information is the same as the image period corresponding to the thermal infrared image sequence; Based on the meteorological description information and the pre-constructed window parameter decision tree, the window parameters are determined, wherein the window parameters include: window length value and window interval value; Construct a sliding window based on the window length value; The sum of the window length value and the window interval value is used as the sliding step size; Based on the sliding step size, the sliding window is controlled to perform sliding windowing on the temperature field image sequence to obtain a temperature field image group sequence.

5. The method for identifying leaks of hazardous chemicals according to claim 4, characterized in that, For each temperature field image group in the temperature field image group sequence, the leakage area is located based on the temperature field image group to generate leakage area information, including: The first temperature field image in the temperature field image group is used to identify the tank boundary to obtain the location of the tank boundary. The boundary deviation is determined based on the tank boundary position and the tank calibration boundary position corresponding to the high-definition infrared camera. In response to the boundary deviation being less than or equal to a preset deviation threshold, a set of search anchor boxes is constructed based on the tank boundary position; For each temperature field image in the temperature field image group, perform the following positioning steps: Global feature extraction is performed on the temperature field image to obtain the temperature field image features; Based on the set of search anchor boxes, feature segmentation is performed on the temperature field image features to obtain a local temperature field image feature set. Parallel leakage area localization is performed based on the local temperature field image feature set to obtain initial leakage area information; Based on the obtained initial set of leakage area information, leakage area information corresponding to the temperature field image group is generated.

6. The method for identifying leaks of hazardous chemicals according to claim 5, characterized in that, The step of filtering the temperature field image sequence based on the temperature field image group corresponding to the target leakage area information to obtain a filtered temperature field image sequence includes: Based on the temperature field image group corresponding to the target leakage area information, determine the starting and ending screening boundary positions; The temperature field image sequence is segmented based on the starting and ending screening boundary positions to obtain a filtered temperature field image sequence.

7. The method for identifying leaks of hazardous chemicals according to claim 6, characterized in that, The step of identifying leakage trend information based on the filtered temperature field image sequence includes: For each filtered temperature field image in the filtered temperature field image sequence, the leakage area of ​​the filtered temperature field image is located to obtain the leakage area coordinates corresponding to the filtered temperature field image in the leakage area coordinate list. Based on the list of leak area coordinates, determine the sequence of leak range areas; Based on the leakage range region sequence, the leakage trend information including the leakage range trend information is determined; Based on the leakage range trend information, a risk level mapping is performed to obtain the leakage risk level included in the leakage trend information.

8. A leak detection device for hazardous chemicals, characterized in that, include: An image correction unit is configured to perform image correction on each thermal infrared image in the thermal infrared image sequence to generate a temperature field image, thereby obtaining a temperature field image sequence, wherein the thermal infrared image sequence is acquired in real time by a high-definition infrared camera facing the chemical storage tank, the high-definition infrared camera covering the long-wave infrared band and the mid-wave infrared band. The discrete windowing unit is configured to perform discrete windowing on the temperature field image sequence to obtain a temperature field image group sequence, wherein a preset number of temperature field images are spaced between every two adjacent temperature field image groups in the temperature field image group sequence. The leakage area localization unit is configured to locate the leakage area for each temperature field image group in the temperature field image group sequence, thereby generating leakage area information, wherein the leakage area information includes: leakage area coordinates and leakage identifier; The image filtering unit is configured to perform image filtering on the temperature field image sequence according to the temperature field image group corresponding to the target leakage area information, to obtain a filtered temperature field image sequence, wherein the target leakage area information includes leakage identifiers that characterize leakage areas where chemical leakage exists; The leakage trend identification unit is configured to identify leakage trends based on the filtered temperature field image sequence to obtain leakage trend information, wherein the leakage trend information includes: a list of leakage area coordinates, leakage range trend information, and leakage risk level.

9. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 7.

10. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images