Method for distinguishing between heat generation and reflected heat in thermal images

Abstract

The method for distinguishing heat generation and reflected heat in a thermal image according to the present invention includes the steps of: a) photographing an object with a thermal imaging camera and collecting a thermal image of the object; b) analyzing characteristics of the thermal image based on temperature distribution; c) detecting heat generation and reflected heat in the thermal image based on the analysis in step b); and d) displaying heat generation and reflected heat areas in the thermal image based on the heat generation and reflected heat in the thermal image detected in step c).

Description

[Technical field]

[0001] The present invention relates to a method for determining heat generation and reflected heat in a thermal image, and more particularly to a method for determining heat generation and reflected heat in a thermal image by photographing and collecting a thermal image from a subject and analyzing the heat generation and reflected heat in the thermal image by artificial intelligence. [Background technology]

[0002] A thermal power plant is a power generation system that generates thermal energy using fossil fuels such as coal and supplies steam generated by the thermal energy to a turbine to produce electricity. A nuclear power plant is a power generation system that generates thermal energy through the nuclear reaction of nuclear fuel and supplies steam generated by the thermal energy to a turbine to produce electricity.

[0003] Since such thermal power plants and nuclear power plants use thermal energy to generate steam to be supplied to turbines, the effects of thermal energy, such as thermal shock, occur in each facility of the power plant or in piping through which high-temperature fluid flows, etc. For example, a nuclear power plant basically includes a reactor where a nuclear reaction occurs and a steam generator that generates steam by the flow of high-temperature primary coolant provided from the reactor, and circulation piping is arranged between the reactor and the steam generator so that the primary coolant circulates.

[0004] Meanwhile, as described above, power plant equipment that uses thermal energy to generate steam requires continuous monitoring of each equipment and piping to prevent damage due to abnormal conditions such as thermal shock caused by thermal energy and to quickly handle malfunctions. For example, a method is used in which thermal images of each equipment and piping of the power plant are collected and the thermal state is analyzed through tracking of high and low temperatures. That is, a diagnostic evaluation is performed depending on the temperature value of the thermal image of each equipment and piping of the power plant, and the analysis of the thermal image is performed by a thermal image analyst.

[0005] However, the method of acquiring temperature values ​​by collecting thermal image data of each equipment and piping of the power plant and tracking high and low temperatures as described above has a problem that the time required for diagnostic evaluation of the diagnostic object increases. Also, diagnostic evaluation based on temperature values ​​obtained by tracking high and low temperatures of thermal image data is performed without accurate discrimination between actual heat generation and reflected heat influence, which reduces the reliability of diagnostic evaluation. In addition, there is a problem that the reliability of diagnostic evaluation varies depending on the knowledge level of the thermal image analyst about the diagnostic object when analyzing the thermal image data. Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a method for determining heat and reflected heat in a thermal image, which can analyze a thermal image of a subject to be diagnosed using artificial intelligence and detect and display heat and reflected heat through the analysis. [Means for solving the problem]

[0007] The above problem is solved by a method for identifying heat generation and reflected heat in a thermal image, comprising the steps of: a) photographing an object with a thermal imaging camera and collecting a thermal image of the object; b) analyzing the characteristics of the thermal image based on the temperature distribution; c) detecting heat generation and reflected heat in the thermal image through the analysis in step b); and d) displaying heat generation and reflected heat areas in the thermal image based on the heat generation and reflected heat detected in step c).

[0008] In the step a), moving image data of a subject is generated by the thermal imaging camera, and the generated moving image data is extracted and collected as the thermal image for each frame.

[0009] The step b) includes: (i) extracting a region of interest from the thermal image through machine learning using an object extraction neural network; (ii) extracting raw data for each pixel of the region of interest; and (iii) extracting temperature data from the region of interest and then managing or filing the extracted temperature data separately.

[0010] The extracted temperature data in the region of interest is refined and regions exhibiting low thermal distribution in the region of interest are searched pixel by pixel for and excluded from the region of interest.

[0011] Heat distributions exhibiting similar characteristics in the region of interest are clustered, and clusters having a size below a certain threshold are removed.

[0012] In the step c), heat spots having a value equal to or greater than a certain threshold are detected in units of pixels in the region of interest.

[0013] In the step c), the temperature change amount between the analysis target pixel and the surrounding pixels in the region of interest is analyzed, and a temperature change amount equal to or greater than a preset threshold is determined to be reflected heat.

[0014] Further details of the embodiments are included in the detailed description and the drawings. Effect of the Invention

[0015] The advantages of the method for distinguishing between heat generation and reflected heat in a thermal image according to the present invention are as follows.

[0016] First, by utilizing artificial intelligence, the temperature distribution of a thermal image of a subject captured and collected by a thermal imaging camera can be analyzed in the field of thermal image diagnostic evaluation, and heat generation and reflected heat of the analyzed thermal image can be detected and displayed, thereby improving the accuracy and reliability of thermal image diagnostic evaluation.

[0017] Secondly, through the heat generation and reflected heat classification algorithm of the area of ​​interest in the thermal image, it is possible to minimize diagnostic errors in real-time thermal image monitoring technology and improve reliability and accuracy, as well as to improve the work efficiency of related workers / operators through visual display of heat generation and reflected heat in the thermal image. [Brief description of the drawings]

[0018] [Figure 1] 1 is a schematic diagram of a system for determining heat generation and reflected heat from a thermal image according to an embodiment of the present invention; [Diagram 2] 4 is an example of a thermal image collected by a method for distinguishing between heat generation and reflected heat in a thermal image according to an embodiment of the present invention; [Diagram 3] 3 is an exemplary diagram showing an example of extracting a region of interest in the thermal image shown in FIG. 2. [Figure 4] FIG. 4 is an example diagram of an extracted region of interest in the thermal image shown in FIG. 3. [Diagram 5] 2 is a flow chart illustrating an operation of a method for determining whether heat is generated or reflected from a thermal image according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a method for determining whether heat is generated or reflected from a thermal image according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

[0020] Prior to the explanation, it should be made clear that the method for distinguishing between heat generation and reflected heat in a thermal image according to an embodiment of the present invention has been described with the subject being limited to power plant equipment that uses thermal energy, but that the subject is not limited thereto and can be applied to equipment in various industrial fields that use thermal energy.

[0021] FIG. 1 is a schematic diagram of a system for distinguishing between heat generation and reflected heat in a thermal image according to an embodiment of the present invention, FIG. 2 is an example diagram of a thermal image collected by a method for distinguishing between heat generation and reflected heat in a thermal image according to an embodiment of the present invention, FIG. 3 is an example diagram of extracting an area of ​​interest in the thermal image shown in FIG. 2, and FIG. 4 is an example diagram of extracting an area of ​​interest in the thermal image shown in FIG. 3.

[0022] 1 to 4, the system 10 for distinguishing heat generation and reflected heat from a thermal image according to an embodiment of the present invention includes a thermal imaging camera 100, a data collector 300, a data analyzer 500, a detector 700, and a display unit 900. The system 10 for distinguishing heat generation and reflected heat from a thermal image according to an embodiment of the present invention distinguishes heat generation and reflected heat from a power plant facility as a subject.

[0023] The thermal imaging camera 100 photographs an object as shown in FIG. 2. Here, the object photographed by the thermal imaging camera 100 includes equipment and piping of a power plant. The thermal imaging camera 100 photographs objects such as equipment and piping of a power plant. The thermal imaging camera 100 photographs a thermal image (F) of the object as a moving image file. The moving image file photographed by the thermal imaging camera 100 is extracted as a thermal image (F) in units of frames of still images. By photographing a moving image file of the object with the thermal imaging camera 100 and extracting the photographed moving image file as a thermal image (F) in units of frames of still images in this way, it is possible to prevent a shortage of learning data when the data analysis unit 500 is trained with the learning data.

[0024] In detail, the thermal imaging camera 100 captures a moving image at 30 frames per second, and the moving image captured by the thermal imaging camera 100 is cut into frames and generated as learning data. When a moving image at 30 frames per second is generated using learning data for each frame as in one embodiment of the present invention, a moving image file of about 1 minute is generated using 1800 frames of learning data.

[0025] The data collection unit 300 collects thermal images (F) captured by the thermal imaging camera 100 as learning data. The data collection unit 300 collects thermal images (F) in units of frames of still images extracted from a moving image file so that the data analysis unit 500 can be trained through machine learning using an object extraction neural network.

[0026] The data analysis unit 500 analyzes the learning data of the thermal image (F) acquired by the data collection unit 300. In detail, the data analysis unit 500 extracts the region of interest (I) from the thermal image (F) through machine learning using an object extraction neural network, extracts pixel-by-pixel raw data from the region of interest (I), extracts temperature data from the region of interest (I), and then manages or files the extracted temperature data separately, as shown in FIG. 3 and FIG. 4. Here, the machine learning using the object extraction neural network used in the data analysis unit 500 is an embodiment of the present invention, and uses a deep learning-based object detection algorithm. As the deep learning-based object detection algorithm, the Faster R-CNN algorithm is used as an embodiment of the present invention, but is not limited thereto, and various deep learning-based object detection algorithms can be used.

[0027] The data analysis unit 500 extracts a region of interest (I) from the thermal image (F) through machine learning using an object extraction neural network, and refines temperature data extracted from the region of interest (I). After refining the temperature data extracted from the region of interest (I), it searches for a region showing low heat distribution in the region of interest (I) in pixel units and excludes it from the region of interest (I).

[0028] Here, refining the temperature data extracted from the region of interest (I) can improve the speed and accuracy of feature analysis of the thermal image (F). After searching and removing areas showing low heat distribution in the region of interest (I) on a pixel-by-pixel basis, heat distributions showing similar characteristics in the region of interest (I) are clustered, and clusters with a size below a certain reference value are excluded. Specifically, during feature analysis of the thermal image (F), since small clusters have a high possibility of measurement error, an exception processing algorithm is applied to exclude clusters below a certain reference value from the analysis, thereby improving the discrimination accuracy of the thermal image (F).

[0029] Next, the detection unit 700 detects heat generation and reflected heat in the thermal image (F) by analyzing the collected thermal image (F). In detail, the detection unit 700 detects heat generation points having a value equal to or greater than a certain threshold in the region of interest (I) in units of pixels, analyzes the temperature change amount between the analyzed pixel and the surrounding pixels in the region of interest (I), and determines that the value equal to or greater than the preset threshold is reflected heat. For example, the top n% heat generation points having a value equal to or greater than a certain threshold in the region of interest (I) are detected in units of pixels. In one embodiment of the present invention, the top 5% heat generation points can be detected in units of pixels. Of course, detecting the top 5% heat generation points in units of pixels is merely one embodiment, and the top n% can be designed differently.

[0030] Finally, the display unit 900 displays the heat generation and reflected heat areas on the thermal image (F) based on the heat generation and reflected heat detected by the detection unit 700. The display unit 900 generates a comparison image of the thermal image (F) and displays the heat generation and reflected heat on the original thermal image (F). Here, although the detection unit 700 and the display unit 900 are separately separated in one embodiment of the present invention, they may be used together.

[0031] FIG. 5 is a flow chart showing an operation of a method for distinguishing between heat generation and reflected heat in a thermal image according to an embodiment of the present invention.

[0032] The method for determining whether a thermal image (F) is a heated or reflected heat according to the embodiment of the present invention is as follows.

[0033] First, a thermal imaging camera 100 captures an object and collects a thermal image (F) of the object (S100). The thermal imaging camera 100 captures a moving image of the power plant's facilities as the object, and the moving image file of the object captured by the thermal imaging camera 100 is extracted as a thermal image (F) in units of still images. In this way, by extracting a thermal image (F) in units of still images from the moving image file captured by the thermal imaging camera 100 and securing it as learning data, it is possible to prevent a shortage of learning data required for machine learning using an object extraction neural network when analyzing the features of the thermal image (F).

[0034] The characteristics of the temperature distribution of the thermal image of the thermal image (F) are analyzed (S300). Here, in step S300, a region of interest (I) of the thermal image (F) is extracted through machine learning using an object extraction neural network, and raw data for each pixel of the region of interest (I) is extracted. Then, in step S300, temperature data of the region of interest (I) is extracted, and the extracted temperature data is separately managed or filed. More specifically, in step S300, the temperature data extracted in the region of interest (I) is refined, and a region showing low heat distribution in the region of interest (I) is searched for in pixel units and excluded from the region of interest (I). Also, in step S300, heat distributions showing similar characteristics in the region of interest (I) are clustered, and clusters having a size less than a certain reference value are excluded. By refining the temperature data extracted from the region of interest (I) and excluding pixels with low heat distribution and excluding heat distribution clusters having a size less than a certain reference value, it is possible to limit the occurrence of errors when detecting heat generation and reflected heat in the thermal image (F), thereby improving the accuracy and reliability of heat generation and reflected heat discrimination in the thermal image (F).

[0035] Heat generation and reflected heat are detected in the thermal image (F) (S500). In step S500, heat generation points having a value equal to or greater than a certain threshold are detected in pixel units in the region of interest (I). In addition, in step S500, the temperature change amount between the analysis target pixel and the surrounding pixels in the region of interest (I) is analyzed, and a value equal to or greater than a preset threshold is determined as reflected heat. More specifically, in step S500, the top n% heat generation points having a value equal to or greater than a certain threshold are detected in pixel units in the region of interest (I). As described above, in one embodiment of the present invention, the top n% heat generation points are described as the top 5% heat generation points, but this is merely an embodiment, and the n% value can be changed according to design changes. Heat generation and reflected heat areas detected in the thermal image (F) are displayed in step S700. In step S700, the heat generation and reflected heat areas in the thermal image (F) detected in step S500 are displayed in the thermal image (F). Specifically, step S700 generates a comparison image of the thermal image (F) to display the heat generation and reflected heat in the original thermal image (F).

[0036] This makes it possible to use artificial intelligence to analyze the temperature distribution of a thermal image of a subject captured and collected by a thermal imaging camera in the field of thermal image diagnostic evaluation, and to detect and display the heat generation and reflected heat of the analyzed thermal image, thereby improving the accuracy and reliability of thermal image diagnostic evaluation.

[0037] In addition, the algorithm for dividing heat and reflected heat in the area of ​​interest in the thermal image can minimize diagnostic errors and improve reliability and accuracy in real-time thermal image monitoring technology, and the visual display of heat and reflected heat in the thermal image can improve the work efficiency of related workers.

[0038] Although the embodiments of the present invention have been described above with reference to the accompanying drawings, those skilled in the art will understand that the present invention can be embodied in other specific forms without changing the technical idea or essential features of the present invention. Therefore, the above described embodiments should be understood to be illustrative and not limiting in all respects. The scope of the present invention is defined by the claims rather than the above detailed description, and all modifications or alterations derived from the meaning and scope of the claims and their equivalents should be interpreted as being included in the scope of the present invention. [Explanation of symbols]

[0039] 100 Thermal Imaging Camera 300 Data Collection Department 500 Data Analysis Department 700 Detector 900 Display section

Claims

1. a) photographing an object with a thermal imaging camera and collecting a thermal image of the object; b) analyzing the thermal image characteristics based on temperature distribution using artificial intelligence; c) utilizing artificial intelligence to detect heat generation and reflected heat in the thermal image analyzed in step b); and d) a step of displaying areas of heat and reflected heat in the thermal image detected in step c) based on the heat and reflected heat in the thermal image.

2. 2. The method of claim 1, wherein in step a), moving image data of the subject is generated by the thermal imaging camera, and the thermal image is extracted and collected by extracting the thermal image frame by frame from the moving image data.

3. The step b) comprises: (i) extracting a region of interest from the thermal image through machine learning using an object extraction neural network; (ii) extracting pixel-by-pixel raw data of the region of interest; and 2. The method of claim 1, further comprising: (iii) extracting the temperature data of the region of interest, and then separately managing or filing the extracted temperature data.

4. 4. The method of claim 3, further comprising: refining the temperature data extracted from the region of interest; searching for an area showing low heat distribution in the region of interest in pixel units; and excluding the area from the region of interest.

5. 4. The method of claim 3, further comprising clustering heat distributions exhibiting similar characteristics in the region of interest, and excluding clusters having a size less than a certain reference value.

6. 4. The method of claim 3, wherein in step c), heat points having a value equal to or greater than a certain threshold value in the region of interest are detected in units of pixels.

7. 4. The method of claim 3, wherein step c) analyzes the temperature change between the analysis target pixel and surrounding pixels in the region of interest, and determines a temperature change above a preset threshold as reflected heat.

Images