White smoke detection device, white smoke detection method and its computer program product

Abstract

A white smoke recognition device, method, and computer program product thereof. The device stores multiple non-overlapping first ranges surrounding a chimney image area and multiple parallel and non-overlapping second ranges, some of which partially overlap with the chimney image area. The device receives multiple original image frames generated by a camera capturing images of the chimney, and subtracts each original image frame from the previous original image frame to generate multiple difference image frames. For each difference image frame, the device determines whether an object satisfies a first condition and / or a second condition based on the difference image frame and the previous difference image frame, thereby determining the white smoke recognition result corresponding to that difference image frame. The first condition is that the object passes through a first subset of the first ranges, and the second condition is that the object passes through a second subset of the second ranges.

Description

[Technical Field]

[0001] This invention relates to a white smoke recognition device, a white smoke recognition method, and a computer program product thereof. Specifically, this invention relates to a white smoke recognition device, a white smoke recognition method, and a computer program product thereof that recognizes the movement of dynamic objects in an image frame. [Previous Technology]

[0002] To comply with environmental regulations, many factories install exhaust gas treatment systems to treat (e.g., denitrification, electrostatic precipitation, desulfurization) flue gas before discharging the treated flue gas into the atmosphere. This treated flue gas contains a large amount of water vapor and is at a temperature higher than that of the atmosphere. Therefore, when it is discharged from the chimney and encounters the cold air, it will condense into supersaturated water vapor, forming white smoke.

[0003] Because white smoke can cause visual disturbances and raise concerns about environmental pollution, some factories install additional equipment to eliminate white smoke emitted from chimneys. For example, some factories install media gas-gas heat exchangers (MGGH), which use pure water as a medium to recover waste heat from the flue gas duct. This recovered waste heat is then used to heat the exhaust gas temperature at the chimney outlet, causing the flue gas to become unsaturated, thereby eliminating white smoke. Under certain climatic conditions (e.g., high temperatures and low humidity), there may still be surplus waste heat after heating the exhaust gas temperature at the chimney outlet. Therefore, the remaining waste heat can be used for other purposes to reduce energy consumption (e.g., heating boiler feedwater to reduce the amount of steam used for heating boiler feedwater). However, under different climatic conditions, the required exhaust gas temperature at the chimney outlet to avoid white smoke varies, making it difficult for operators to determine the optimal temperature. Especially when the weather or unit load changes, operators often cannot immediately adjust the flue gas temperature at the chimney outlet to the optimal temperature point, resulting in wasted heat energy or the generation of white smoke.

[0004] In view of this, there is an urgent need in the field for a white smoke identification technology that can monitor whether a factory is emitting white smoke, and an operational recommendation mechanism that can maximize the benefits of waste heat recovery when the factory does not produce white smoke from the chimney. [Summary of the Invention]

[0005] One object of the present invention is to provide a white smoke recognition device. The white smoke recognition device includes a storage device, a transceiver interface, and a processor, wherein the processor is electrically connected to the storage device and the transceiver interface. The storage device stores a plurality of first ranges and a plurality of second ranges defined according to an image frame range, wherein the first ranges do not overlap with each other and surround a chimney image range, and the second ranges are parallel to each other and do not overlap, with some of the second ranges partially overlapping the chimney image range. The transceiver interface receives a plurality of original image frames in chronological order generated by a camera capturing images of a chimney. The processor subtracts each of the original image frames from its corresponding preceding original image frame to generate a plurality of difference image frames, and performs the following operation for each difference image frame: determining, based on the difference image frame and its corresponding preceding difference image frame, whether an object satisfies one of a first condition and a second condition, thereby determining a white smoke recognition result corresponding to the difference image frame. The first condition is that the object passes through a first subset of the first ranges, and the second condition is that the object passes through a second subset of the second ranges.

[0006] Another object of the present invention is to provide a white smoke identification method, which is executed by an electronic computing device. The electronic computing device stores a plurality of first ranges and a plurality of second ranges defined according to an image frame range, wherein the first ranges do not overlap with each other and surround a chimney image range, the second ranges are parallel to each other and do not overlap, and a portion of the second ranges partially overlaps with the chimney image range. The white smoke identification method includes the following steps: (a) receiving a plurality of original image frames in chronological order generated by a camera capturing images of a chimney; (b) subtracting each of the original image frames from the corresponding previous original image frame to generate a plurality of difference image frames; and (c) performing the following steps for each difference image frame: (c1) determining, based on the difference image frame and the corresponding previous difference image frame, whether there exists an object that satisfies one of a first condition and a second condition, thereby determining a white smoke identification result corresponding to the difference image frame, wherein the first condition is that the object passes through a first subset of the first ranges, and the second condition is that the object passes through a second subset of the second ranges.

[0007] Another object of the present invention is to provide a computer program product. After an electronic computing device loads the computer program product, the electronic computing device executes a plurality of program instructions contained in the computer program product to implement the white smoke recognition method as described above.

[0008] The white smoke recognition technology (including at least an apparatus, method, and computer program product) provided by the present invention separates dynamic objects from static backgrounds by subtracting each original image frame from the corresponding previous original image frame, making the obtained plurality of difference image frames more clearly present dynamic objects (including white smoke emitted from the chimney) and excluding static backgrounds (including stationary white clouds). Furthermore, the white smoke recognition technology provided by the present invention identifies whether an object moves upward, to the side, or to the upper side from the chimney opening by defining a plurality of non-overlapping first ranges surrounding a chimney image range and determining whether an object passes through a subset of these first ranges in consecutive difference image frames. Additionally, the white smoke recognition technology provided by the present invention identifies whether an object moves downward from the chimney opening by defining a plurality of parallel and non-overlapping second ranges (partially overlapping with the chimney image range) and determining whether an object passes through a subset of these second ranges in consecutive difference image frames. Therefore, regardless of weather conditions (e.g., daytime, nighttime, cloudy, clear, wind direction), the white smoke identification technology provided by this invention can accurately identify whether a factory chimney is emitting white smoke.

[0009] The following detailed description of the technology and implementation of the present invention is illustrated in conjunction with the drawings, so that those skilled in the art can understand the technical features of the claimed invention.

Implementation Method

[0010] The following will explain the white smoke recognition device, white smoke recognition method, and computer program product provided by the present invention through embodiments. However, these embodiments are not intended to limit the implementation of the present invention to any environment, application, or manner described in these embodiments. The description of the following embodiments is only for illustrating the purpose of the present invention and is not intended to limit the scope of the present invention. It should be understood that in the following embodiments and drawings, elements not directly related to the technical features of the present invention have been omitted and not described or / and illustrated. In addition, the dimensions of the elements and the proportional relationships between the elements in the drawings are only for illustration and explanation and are not intended to limit the scope of the present invention. Furthermore, unless otherwise stated, the terms "a," "the," and similar terms used in this specification and the claims should be understood to include both singular and plural forms.

[0011] Figure 1 depicts a schematic diagram of the architecture of a white smoke detection device 1 according to some embodiments of the present invention. The white smoke detection device 1 includes a storage device 11, a processor 13, and a transceiver interface 15, and the processor 13 is electrically connected to the storage device 11 and the transceiver interface 15. The storage device 11 may include one or more of the following: memory, hard disk, and any other non-transitory storage medium, circuit, or device with the same function known to those skilled in the art to which this invention pertains. The processor 13 may include one or more of the following: various processors, central processing units (CPUs), microprocessors (MPUs), digital signal processors (DSPs), graphics processing units (GPUs), and other computing devices known to those skilled in the art to which this invention pertains. The transceiver interface 15 may be a wired interface or a wireless interface, which may be configured to be electrically connected to a network (e.g., an Internet, a local area network) to receive data from other electronic devices on the network, or may be configured to be electrically connected to an electronic device to receive data directly from that electronic device.

[0012] White smoke identification technology

[0013] To enable the white smoke detection device 1 to identify whether a factory chimney is emitting white smoke, a camera (not shown) needs to be fixedly mounted near the chimney, and the camera's field of view needs to cover the chimney and its surroundings (e.g., covering both sides and a predetermined distance above the chimney). As shown in Figure 2A, the image frame generated by the camera has an image frame range IR, and the image frame range IR has a chimney image range FR (i.e., the range occupied by the chimney in the image frame as captured by the camera). The present invention defines a plurality of first ranges and a plurality of second ranges based on the image frame range IR and the chimney image range FR, wherein the first ranges do not overlap with each other and surround the chimney image range FR, the second ranges are parallel to each other and do not overlap, and a portion of the second ranges partially overlaps with the chimney image range FR.

[0014] For ease of understanding and explanation, the following will describe in detail a plurality of first ranges using the specific example shown in Figure 2B, and the following will describe in detail a plurality of second ranges using the specific example shown in Figure 2C. However, these specific examples are not intended to limit the scope of the present invention. In the specific example shown in Figure 2B, six first ranges A1, A2, A3, A4, A5, and A6 are defined within the image frame range IR. However, it should be understood that the present invention does not limit the number of first ranges to six. As shown in Figure 2B, the first ranges A1 to A6 are all in the shape of the phonetic symbol ㄇ, and the first ranges A1 to A6 surround the chimney image range FR, and the first ranges A1 to A6 do not overlap with each other. In the specific example shown in Figure 2C, eight second ranges B1, B2, B3, B4, B5, B6, B7, and B8 are defined within the image frame range IR. However, it should be understood that the present invention does not limit the number of second ranges to eight. As shown in Figure 2C, the second ranges B1 to B8 are parallel to each other and do not overlap, and a portion of these second ranges (i.e., second ranges B3, B4, B5, B6, B7, B8) partially overlaps with the chimney image range FR.

[0015] It should be noted that the present invention does not limit how the chimney image range FR, the first range A1-A6, and the second range B1-B8 are determined, as long as the determined first range A1-A6 and the second range B1-B8 meet the aforementioned conditions. In some embodiments, the white smoke identification device 1 may first receive a reference image frame ID (e.g., an image frame captured by a camera when the factory chimney is not emitting white smoke (e.g., when the factory is not operating) via the transceiver interface 15), and then the processor 13 may use image recognition technology to identify the chimney in the reference image frame ID as the chimney image range FR, and the processor 13 may define the first range A1-A6 and the second range B1-B8 within the image frame range IR. In some embodiments, the user may mark the chimney image range FR, the first range A1-A6, and the second range B1-B8 based on the content of the reference image frame ID.

[0016] The storage device 11 stores the first range A1 to A6 and the second range B1 to B8. It should be noted that the present invention does not limit the form in which the storage device 11 stores the first range A1 to A6 and the second range B1 to B8. In some embodiments, the storage device 11 may store multiple coordinate values ​​of each of the first range A1 to A6 and the second range B1 to B8 in the image frame range IR.

[0017] To identify whether a factory chimney is emitting white smoke, the transceiver interface 15 receives a plurality of original image frames I1, I2, I3, ..., In-1, In, generated by a camera capturing images of the chimney in chronological order. In some embodiments, the transceiver interface 15 is electrically connected to a network and receives the original image frames I1, I2, I3, ..., In-1, In from the camera via the network. In some embodiments, the transceiver interface 15 is electrically connected to the camera and receives the original image frames I1, I2, I3, ..., In-1, In directly from the camera. Furthermore, in some embodiments, the transceiver interface 15 receives the original image frames I1, I2, I3, ..., In-1, In in real time. In some embodiments, the transceiver interface 15 does not receive the original image frames I1, I2, I3, ..., In-1, In in real time.

[0018] For each of the original image frames I1, I2, I3, ..., In-1, In, if it has a corresponding previous original image frame, the processor 13 subtracts it from the corresponding previous original image frame to generate a plurality of difference image frames D1, D2, ..., Dn-1. As shown in Figure 3, the processor 13 subtracts the corresponding previous original image frame I1 from the original image frame I2 to generate difference image frame D1, subtracts the corresponding previous original image frame I2 from the original image frame I3 to generate difference image frame D2, ..., and subtracts the corresponding previous original image frame In-1 from the original image frame In to generate difference image frame Dn-1. By subtracting each original image frame from the corresponding previous original image frame, dynamic objects (including white smoke emitted by the chimney) and static backgrounds (including stationary white clouds) can be separated, making the difference image frames D1 to Dn-1 more clearly present the dynamic objects. Since the white smoke emitted from chimneys is a moving, dynamic object, while the white clouds in the sky are mostly static, by subtracting each original image frame from the corresponding previous original image frame, the difference image frames D1 to Dn-1 can more clearly show the moving white smoke and exclude the static white clouds, making subsequent identification more accurate.

[0019] For each of the difference image frames D1 to Dn-1, the processor 13 performs the following operation: based on the difference image frame and the corresponding previous difference image frame, it determines whether there is an object that satisfies at least one of the first condition and the second condition, thereby determining the white smoke identification result corresponding to the difference image frame. The first condition is that the object passes through a subset of the first range A1 to A6, and the second condition is that the object passes through a subset of the second range B1 to B8. The first condition is used to determine whether there is an object moving upwards, to the side, or to the upper side from the location of the chimney (as shown by the multiple arrows in Figure 2B), and these directions are the usual directions of movement of white smoke. The second condition is used to determine whether there is an object moving downwards from the location of the chimney (as shown by the multiple arrows in Figure 2C), and this direction is the direction of movement of white smoke under specific weather conditions (e.g., northeast wind). When the processor 13 determines, based on the difference image frame and the corresponding previous difference image frame, that an object satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is "white smoke detected". When the processor 13 determines, based on the difference image frame and the corresponding previous difference image frame, that no object satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is "white smoke not detected".

[0020] For example, if the processor 13 determines, based on the difference image frame D2 and the corresponding previous difference image frame D1, that an object passes through a subset of the first range A1 to A6 (e.g., sequentially passing through the first range A2, A3, A4), then the white smoke recognition result corresponding to the difference image frame D2 is that white smoke has been detected. As another example, if the processor 13 determines, based on the difference image frame Dn-2 and the corresponding previous difference image frame (not shown), that an object passes through a subset of the second range B1 to B8 (e.g., sequentially passing through the second range B3, B4, B5, B6), then the white smoke recognition result corresponding to the difference image frame Dn-2 is that white smoke has been detected. For another example, if the processor 13 determines, based on the difference image frame Dn-1 and the corresponding previous difference image frame Dn-2, that no object passes through a subset of the first range A1 to A6 and no object passes through a subset of the second range B1 to B8, then the white smoke recognition result corresponding to the difference image frame Dn-1 is that no white smoke was recognized.

[0021] As described above, the white smoke recognition device 1 separates dynamic objects from static backgrounds by subtracting each original image frame from the corresponding previous original image frame, making the obtained difference image frames more clearly present dynamic objects (including white smoke emitted from the chimney) and excluding static backgrounds (including stationary white clouds). Furthermore, the white smoke recognition device 1 identifies whether an object is moving upwards, to the side, or to the upper side from the chimney opening by defining a plurality of non-overlapping first ranges surrounding the chimney image range FR and determining whether an object passes through a subset of these first ranges in consecutive difference image frames. Since these first ranges surround the chimney image range FR, and upwards, to the side, or to the upper side are the usual directions of movement for white smoke from the chimney outlet, the aforementioned mechanism can accurately identify white smoke moving in the usual direction of movement. Furthermore, the white smoke identification device 1 identifies whether an object is moving downwards from the chimney opening by defining a plurality of parallel and non-overlapping second ranges (parts of which overlap with the chimney image range FR) and determining whether an object passes through a subset of these second ranges in consecutive difference image frames. Since parts of these second ranges overlap with the chimney image range FR, and downward movement from the chimney opening is the direction of white smoke movement under specific weather conditions (e.g., northeasterly winds), the aforementioned mechanism can accurately identify white smoke emitted by the chimney under specific weather conditions. Therefore, regardless of weather conditions (e.g., daytime, nighttime, cloudy, clear, wind direction), the white smoke identification device 1 can accurately identify whether a factory chimney is emitting white smoke.

[0022] Establishment of White Smoke Prediction Model M

[0023] In some embodiments, the white smoke identification device 1 may further utilize the aforementioned white smoke identification results to establish a white smoke prediction model M for the factory. In these embodiments, it is necessary to first determine a plurality of characteristic variables of the factory equipment, and then establish the white smoke prediction model M based on the values ​​corresponding to these characteristic variables in a plurality of historical data sets and the white smoke identification results.

[0024] The method for determining the plurality of characteristic variables of the factory equipment will now be explained. As shown in Figure 4, in some embodiments, the storage 11 stores a plurality of historical data sets H11, ..., H1N from the past operation of the factory. Each record in the historical data set H11, ..., H1N contains a plurality of numerical values ​​(not shown) that correspond one-to-one with a plurality of given variables (not shown) of the factory equipment. The processor 13 analyzes the historical data sets H11, ..., H1N using a forward feature selection method (not shown) to select a plurality of characteristic variables (not shown) from these given variables. It should be noted that those skilled in the art to which this invention pertains will understand the technical details of the forward feature selection method, and therefore will not be elaborated upon. The characteristic variables selected by the forward feature selection method are the more important of the given variables related to the operation of the factory equipment.

[0025] For example, if a plant is equipped with a heat exchanger and wants to know the operating status of the heat exchanger, the given variables may include many of the following: main steam flow rate, chimney outlet temperature, wet bulb temperature, dry bulb temperature, inlet flue gas temperature of the heat exchanger's heat exchanger, outlet flue gas temperature of the heat exchanger's heat exchanger, inlet water temperature of the heat exchanger's heat exchanger, inlet flue gas temperature of the heat exchanger's reheater, inlet water temperature of the heat exchanger's reheater, condensate intake water temperature, condensate return water temperature, circulating water flow rate of the heat exchanger, feedwater heater flow rate, and other variables. These characteristic variables are a subset of the given variables, such as: chimney outlet temperature, wet-bulb temperature, dry-bulb temperature, inlet flue gas temperature of the heat exchanger of the heat exchanger, outlet flue gas temperature of the heat exchanger of the heat exchanger, inlet flue gas temperature of the reheater of the heat exchanger, condensate intake water temperature, and condensate return water temperature.

[0026] In some embodiments, the characteristic variables can be selected in another way. For example, process professionals can first select a plurality of candidate variables from the given variables based on their expertise, and then the processor 13 can analyze the historical data group H11, ..., H1N using a forward feature selection method to select the characteristic variables from the candidate variables. In some embodiments, if computational efficiency and / or overfitting issues are not considered, all the given variables can also be used as characteristic variables.

[0027] After confirming the characteristic variables, a white smoke prediction model M can be constructed, and its technical details are explained below. As shown in Figure 4, in some embodiments, the storage device 11 stores multiple historical data sets H21, ..., H2M from the past operation of the factory. Each record in the historical data sets H21, ..., H2M contains a white smoke identification result at a certain point in time and multiple values ​​corresponding one-to-one with these characteristic variables. It should be noted that if the aforementioned historical data sets H11, ..., H1N also each contain a white smoke identification result, the white smoke prediction model M can also be constructed using the historical data sets H11, ..., H1N.

[0028] The historical data set H21, ..., H2M is divided into a first subset and a second subset, and the first subset and the second subset have no overlap. The processor 13 trains a machine learning model (e.g., a Rogers regression model) based on the first subset of the historical data set H21, ..., H2M to obtain a preliminary white smoke prediction model M. It should be noted that the white smoke prediction model M is designed so that the input corresponds to the feature variables, and the output corresponds to the white smoke recognition result. Those skilled in the art will understand how the processor 13 trains a machine learning model (e.g., a Rogers regression model) based on the first subset of the historical data set H21, ..., H2M, which will not be elaborated upon here.

[0029] The processor 13 further verifies whether the preliminary white smoke prediction model M meets the requirements based on the second subset of historical data sets H21, ..., H2M. In some embodiments, the processor 13 may use a confusion matrix to record the prediction results of the white smoke prediction model M for the second subset of historical data sets H21, ..., H2M, as shown in Table 1: Confusion Matrix Actual situation smokeless There is smoke White smoke prediction model M prediction smokeless N 11 N 12 There is smoke N 21 N 22 Table 1

[0030] In Table 1, parameter N11 represents the number of chimneys that actually did not emit white smoke but were predicted not to emit by the white smoke prediction model M; parameter N12 represents the number of chimneys that actually emitted white smoke but were predicted not to emit by the white smoke prediction model M; parameter N21 represents the number of chimneys that actually did not emit white smoke but were predicted to emit by the white smoke prediction model M; and parameter N22 represents the number of chimneys that actually emitted white smoke and were predicted to emit by the white smoke prediction model M. Processor 13 calculates a precision rate based on the values ​​in Table 1, which is the precision rate at which the white smoke prediction model M predicts no smoke. The value is P, where P represents precision. Processor 13 also calculates a recall rate based on the values ​​in Table 1, which is the ratio of chimneys that actually emitted no smoke but were predicted by the white smoke prediction model M. The value is R, where R represents recall. Processor 13 then calculates an F1 score based on the precision and recall rate. The value is F, where F represents F1 score. If the F1 score is greater than a preset threshold (e.g., 0.85), the white smoke prediction model M has passed validation (meaning the prediction results of the white smoke prediction model M are accurate enough) and can therefore be deployed online. If the F1 score is not greater than the preset threshold, training and validation need to be performed again. The aforementioned training and validation can be repeated multiple times until the white smoke prediction model M passes validation.

[0031] Application of White Smoke Prediction Model M

[0032] In some embodiments, the white smoke detection device 1 may utilize a validated white smoke prediction model M to provide operational recommendations that prevent the factory's chimney from emitting white smoke during operation. The operational recommendations provided by the white smoke detection device 1 relate to the temperature that should be reached at the chimney outlet.

[0033] Specifically, when the factory equipment is operating, the values ​​of these characteristic variables of the factory equipment should be set to fixed or controllable. In some embodiments, the value of the chimney outlet temperature among these characteristic variables is controllable, while the values ​​of the other characteristic variables are known and fixed. In these embodiments, the processor 13 further inputs a plurality of preset chimney outlet temperatures T1, ..., TK and a real-time data set DS (i.e., the values ​​corresponding to the other characteristic variables) into the white smoke prediction model M to obtain a plurality of white smoke prediction results PR1, ..., PRK, as shown in the data flow diagram in Figure 5. Each white smoke prediction result is either a prediction of smoke or a prediction of no smoke. The processor 13 then selects the lowest preset chimney outlet temperature from the white smoke prediction results PR1, ..., PRK that are predicted to be smokeless as a suggested chimney outlet temperature. Taking the specific example shown in Table 2 as an example, the processor 13 will select the lowest one (i.e., 72 degrees Celsius) from the preset chimney outlet temperatures (i.e., 72 degrees Celsius, 73 degrees Celsius, 74 degrees Celsius and 75 degrees Celsius) corresponding to the white smoke prediction result of no smoke as a suggested chimney outlet temperature. Preset chimney outlet temperature White smoke forecast results 66 degrees Celsius There is smoke 67 degrees Celsius There is smoke 68 degrees Celsius There is smoke 69 degrees Celsius There is smoke 70 degrees Celsius There is smoke 71 degrees Celsius There is smoke 72 degrees Celsius smokeless 73 degrees Celsius smokeless 74 degrees Celsius smokeless 75 degrees Celsius smokeless Table 2

[0034] Through the aforementioned operating mechanism, regardless of climatic conditions, the white smoke detection device 1 can utilize the validated white smoke prediction model M to provide the lowest chimney outlet temperature at which the chimney will not emit white smoke during the operation of the factory equipment, and the factory equipment can then adopt this temperature. For example, if the factory installs a heat exchanger to recover waste heat from the flue, the recovered waste heat can be used to heat the chimney outlet to the lowest chimney outlet temperature recommended by the white smoke detection device 1, and the remaining waste heat can be used for other purposes, maximizing the efficiency of the recovered waste heat (i.e., using the least amount of waste heat to prevent the chimney from emitting white smoke, so as to retain more waste heat for other purposes).

[0035] Methods and Applications for Identifying White Smoke

[0036] The present invention also provides a white smoke recognition method, which is executed by an electronic computing device (e.g., white smoke recognition device 1). The electronic computing device stores a plurality of first ranges and a plurality of second ranges defined according to an image frame range, wherein the first ranges do not overlap with each other and surround a chimney image range, the second ranges are parallel to each other and do not overlap, and some of the second ranges partially overlap with the chimney image range. The main flowchart of the white smoke recognition method is depicted in Figure 6, which includes steps S601, S603 and S605.

[0037] In step S601, the electronic computing device receives a plurality of original image frames in chronological order generated by a camera capturing images of a chimney. In step S603, the electronic computing device subtracts each of the original image frames from the corresponding previous original image frame to generate a plurality of difference image frames.

[0038] In step S605, the following steps are performed for each difference image frame: Based on the difference image frame and the corresponding previous difference image frame, it is determined whether an object satisfies at least one of a first condition and a second condition, thereby determining the white smoke recognition result corresponding to the difference image frame. The first condition is that the object passes through a first subset of the first ranges, and the second condition is that the object passes through a second subset of the second ranges. If, based on the difference image frame and the corresponding previous difference image frame, it is determined that an object satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is that white smoke has been detected. If, based on the difference image frame and the corresponding previous difference image frame, it is determined that no object satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is that white smoke has not been detected.

[0039] In some embodiments, the electronic computing device further stores a plurality of first historical data sets of a factory, each of the first historical data sets containing a white smoke identification result at a certain point in time and a plurality of values ​​corresponding one-to-one with a plurality of feature variables. These feature variables are the more important of a plurality of given variables related to the operation of the factory equipment. In these embodiments, the white smoke identification method further includes a step of training a machine learning model based on the first historical data sets to obtain a white smoke prediction model.

[0040] In some embodiments, the white smoke identification method further includes a step of inputting a plurality of preset chimney outlet temperatures and a set of real-time data into the white smoke prediction model to obtain a plurality of white smoke prediction results, wherein each white smoke prediction result is either a prediction of smoke or a prediction of no smoke. The white smoke identification method further includes another step of selecting the smallest preset chimney outlet temperature from the white smoke prediction results that predict no smoke as an actual chimney outlet temperature.

[0041] In some embodiments, the electronic computing device further stores a plurality of second historical data sets of the factory, wherein each of the second historical data sets contains a plurality of values ​​corresponding one-to-one with the given variables. The white smoke identification method further includes a step of analyzing the second historical data sets using a forward feature selection method to select the feature variables from the given variables.

[0042] In addition to the above steps, the white smoke identification method provided by the present invention can also perform other steps to have the functions of the white smoke identification device 1 in the aforementioned embodiments and achieve the same technical effect. Those skilled in the art can directly understand how the white smoke identification method provided by the present invention performs these steps based on the aforementioned embodiments, has the same function, and achieves the same technical effect, so it will not be described in detail.

[0043] The white smoke recognition method described in the above embodiments can be implemented by a computer program product containing a plurality of program instructions. The computer program product can be a file that can be transmitted over a network, or it can be stored in a non-transitory computer-readable storage medium. After the program instructions contained in the computer program product are loaded into an electronic computing device (e.g., white smoke recognition device 1), the computer program executes the white smoke recognition method as described in the above embodiments. The non-transitory computer-readable storage medium can be an electronic product, such as: a read-only memory (ROM), a flash memory, a floppy disk, a hard disk, a compact disk (CD), a digital versatile disc (DVD), a USB flash drive, a database accessible via a network, or any other storage medium known to those skilled in the art and having the same function.

[0044] It should be noted that certain terms in the specification and claims of this invention (including at least: scope, conditions, subset, historical data set) are preceded by "first" or "second". These "first" and "second" are used to distinguish these terms from each other. Unless otherwise specified, or if the order of these terms is not apparent from the context, the order of these terms is not restricted by the prefix "first" or "second".

[0045] In summary, the white smoke identification technology (including at least a device, method, and computer program product) provided by the present invention generates differential image frames that more clearly present dynamic objects and determines whether an object passes through a subset of a plurality of first ranges and / or a subset of a plurality of second ranges in consecutive differential image frames. This enables accurate identification of whether a factory chimney emits white smoke under various weather conditions (e.g., daytime, nighttime, cloudy, clear, wind direction). The white smoke identification technology provided by the present invention can further establish a white smoke prediction model and utilize this model to determine the lowest chimney outlet temperature at which the factory equipment will not emit white smoke during operation. This temperature is then used by the factory equipment to maximize the efficiency of recovered waste heat (i.e., using minimal waste heat to prevent white smoke emission from the chimney, thus retaining more waste heat for other uses).

[0046] The above embodiments are used to illustrate some aspects of the present invention and explain the technical features of the present invention, and are not intended to limit the scope and range of protection of the present invention. Any changes or equivalent arrangements that can be easily made by those skilled in the art to which this invention pertains are within the scope of the present invention, and the scope of protection of the present invention is determined by the claims. [Simplified Explanation of the Diagram]

[0047] Figure 1 depicts a schematic diagram of the architecture of a white smoke detection device in some embodiments.

[0048] Figure 2A depicts a specific example of the range of an image frame.

[0049] Figure 2B depicts a specific example of a plurality of first ranges.

[0050] Figure 2C depicts a specific example of a plurality of second ranges.

[0051] Figure 3 is a schematic diagram of generating multiple difference image frames using multiple original image frames.

[0052] Figure 4 depicts a schematic diagram of the architecture of a white smoke detection device in some other embodiments.

[0053] Figure 5 depicts a data flow diagram for white smoke prediction using a white smoke prediction model.

[0054] Figure 6 depicts a main flowchart of the white smoke identification method in some embodiments.

Claims

1. A white smoke recognition device, comprising: a storage unit storing a plurality of first ranges and a plurality of second ranges defined by an image frame range, wherein the first ranges do not overlap and surround a chimney image range, the second ranges are parallel to each other and do not overlap, and a portion of the second ranges partially overlaps with the chimney image range; a transceiver interface receiving a plurality of original image frames in chronological order generated by a camera capturing images of a chimney; and a processor electrically connected to the storage unit and the transceiver interface, subtracting each of the original image frames from a corresponding preceding original image frame to generate a plurality of difference image frames, and performing the following operations for each difference image frame: determining, based on the difference image frame and the corresponding preceding difference image frame, whether an object satisfies at least one of a first condition and a second condition, thereby determining a white smoke recognition result corresponding to the difference image frame, wherein... The first condition is that the object passes through a first subset of the first range, and the second condition is that the object passes through a second subset of the second range. The storage device further stores a plurality of first historical data sets of a factory, each first historical data set containing one of the white smoke identification results and a plurality of numerical values ​​corresponding one-to-one with a plurality of feature variables. The processor further performs the following operations: training a machine learning model based on a first subset of the first historical data sets to obtain a white smoke prediction model; validating the white smoke prediction model based on a second subset of the first historical data sets; inputting a plurality of preset chimney outlet temperatures and a real-time data set into the white smoke prediction model to obtain a plurality of white smoke prediction results, each white smoke prediction result being either a prediction of smoke or a prediction of no smoke; and selecting the lowest preset chimney outlet temperature from the white smoke prediction results that predict no smoke as a suggested chimney outlet temperature.

2. The white smoke recognition device as claimed in claim 1, wherein when the processor determines, based on the difference image frame and the corresponding previous difference image frame, that there exists an object that satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is that white smoke has been recognized.

3. The white smoke recognition device as claimed in claim 1, wherein when the processor determines, based on the difference image frame and the corresponding previous difference image frame, that there is no object that satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is that no white smoke was recognized.

4. The white smoke detection device as claimed in any one of claims 1 to 3, wherein the storage further stores a plurality of second historical data sets of the factory, each of the second historical data sets containing a plurality of values ​​corresponding one-to-one with a plurality of given variables, wherein, The processor further analyzes the second set of historical data using a forward feature selection method to select the feature variables from the given variables.

5. A white smoke identification method, performed by an electronic computing device storing a plurality of first ranges and a plurality of second ranges defined according to an image frame range, the first ranges being non-overlapping and surrounding a chimney image range, the second ranges being parallel and non-overlapping, and portions of the second ranges partially overlapping the chimney image range, the white smoke identification method comprising the following steps: (a) receiving a plurality of original image frames in chronological order generated by a camera capturing images of a chimney; (b) subtracting each of the original image frames from its corresponding preceding original image frame to generate a plurality of difference image frames; and (c) performing the following steps for each of the difference image frames: (c1) Based on the difference image frame and the corresponding previous difference image frame, determine whether there is an object that satisfies at least one of a first condition and a second condition, thereby determining the white smoke recognition result corresponding to the difference image frame, wherein the first condition is that the object passes through a first subset of the first ranges, and the second condition is that the object passes through a second subset of the second ranges; The electronic computing device further stores a plurality of first historical data sets of a factory, each of the first historical data sets containing one of the white smoke identification results and a plurality of values ​​corresponding one-to-one with a plurality of feature variables, and the white smoke identification method further includes the following steps: (d) training a machine learning model based on a first subset of the first historical data sets to obtain a white smoke prediction model; (e) validating the white smoke prediction model based on a second subset of the first historical data sets; (f) inputting a plurality of preset chimney outlet temperatures and a real-time data set into the white smoke prediction model to obtain a plurality of white smoke prediction results, wherein each white smoke prediction result is either a prediction of smoke or a prediction of no smoke; and (g) selecting the lowest preset chimney outlet temperature from the white smoke prediction results that predict no smoke as a suggested chimney outlet temperature.

6. The white smoke recognition method as described in claim 5, wherein when step (c1) determines, based on the difference image frame and the corresponding previous difference image frame, that there exists an object that satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is white smoke recognized.

7. The white smoke recognition method as described in claim 5, wherein when step (c1) determines, based on the difference image frame and the corresponding previous difference image frame, that there is no object that satisfies at least one of the first condition and the second condition, the white smoke recognition result corresponding to the difference image frame is that no white smoke was recognized.

8. The white smoke identification method according to any one of claims 5 to 7, wherein the electronic computing device further stores a plurality of second historical data sets of the factory, each of the second historical data sets containing a plurality of values ​​corresponding one-to-one with a plurality of given variables, the white smoke identification method further comprising the following steps: analyzing the second historical data sets using a forward feature selection method to select the feature variables from the given variables.

9. A computer program product, after being loaded via an electronic computing device, wherein the electronic computing device executes a plurality of program instructions contained in the computer program product to implement the white smoke recognition method as described in any one of claims 5 to 8.

Images