System for object recognition detection of nuclear power plant control logic text and method therefor

Abstract

The present invention relates to an object recognition detection system for nuclear power plant control logic characters, and more specifically, to an object recognition detection system and method for nuclear power plant control logic characters to prevent false positives in drawing symbol recognition. The present invention comprises a collection unit that collects data on the size and coordinate values ​​of each symbol in a nuclear power plant control logic drawing; a setting unit that recognizes character symbols as individual objects based on lines within the drawing, performs data learning through image recognition of character symbols as single objects, and sets character symbols with noise as a single image based on the learned data; a detection unit that detects objects by determining the recognizable size range of each symbol and setting a grace period for the symbol range; and a recognition unit that recognizes symbols during drawing recognition by setting a range value that yields high accuracy for each symbol, thereby having the effect of reducing the false positive rate of reading symbols as characters.

Description

Technology Field

[0001] The present invention relates to an object recognition detection system for nuclear power plant control logic characters, and more specifically, to an object recognition detection system for nuclear power plant control logic characters and a method thereof to prevent false positives in drawing symbol recognition. Background Technology

[0002] False positives due to the size of symbols in nuclear power plant drawings refer to the problem where important elements necessary for operation are omitted because the size of parts or equipment indicated on the drawing is too small, or where errors occur in the interpretation of the drawing because the size is indicated too large.

[0003] The size of symbols is critical in nuclear power plant drawings. The problem of false positives caused by symbol size in nuclear power plant drawings can be broadly categorized into two types. First, cases where symbols are too small to be read properly, and second, cases where symbols are too large or disproportionate, leading to misunderstandings.

[0004] Regarding the first issue, the symbols for parts or equipment are displayed in small sizes on the drawings, which increases the likelihood of misunderstanding among engineers or workers interpreting them. This can be a particularly serious problem for small parts or equipment. For instance, if the symbols for small parts or machinery operating inside a reactor are displayed in small sizes, failure to properly identify them could pose a significant risk to reactor safety.

[0005] The second issue involves cases where symbols are too large or disproportionate, leading to misunderstandings. For instance, using a large symbol for a small part exaggerates its size, potentially resulting in a discrepancy with its actual dimensions. This compromises design accuracy and can cause problems during manufacturing or repair processes.

[0006] Accordingly, although drawings of nuclear power plants are generated according to a set standard, false positives may occur because the size of each symbol varies, and the size difference between the smallest and largest symbols can be up to 10 times, so the recognizable area for each symbol is not defined, leading to false positives.

[0007] In addition, the control logic drawings of nuclear power plants have a lot of noise, so when recognized using the character recognition tool Tesseract, they show a low recognition rate of 65.32%. Prior art literature

[0008] Registered Patent 10-2104605 (2020.04.20.) The problem to be solved

[0009] The present invention aims to solve the aforementioned problem by providing an object recognition detection system for nuclear power plant control logic characters to reduce false positives by setting recognizable areas for each symbol when using an artificial intelligence-based control logic drawing verification tool, thereby preventing false positives in drawing symbol recognition. means of solving the problem

[0010] To achieve the above objective, the present invention comprises: a collection unit that collects data on the size and coordinate values ​​of each symbol in a nuclear power plant control logic drawing; a setting unit that recognizes character symbols as individual objects based on lines within the drawing, performs data learning through image recognition of the character symbols as a single object unit, and sets the character symbols with noise as a single image based on the learned data; a detection unit that detects objects by determining the recognizable size range of each symbol and setting a grace period for the symbol range; and a recognition unit that recognizes symbols during drawing recognition by setting a range value that yields high accuracy for each symbol. Effects of the invention

[0011] According to the present invention, when using an AI-based control logic drawing verification tool to prevent false positives in drawing symbol recognition, a recognizable area for each symbol is set to prevent false positives, thereby improving the recognition rate. Additionally, it reduces the false positive rate of reading symbols as text. Brief explanation of the drawing

[0012] FIG. 1 is a configuration diagram showing an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention. FIG. 2 is a diagram showing the recognition of characters as individual objects through the setting unit of an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention. FIG. 3 is a diagram showing individually recognized learning data through the setting unit of an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention. Figure 4 shows the result of recognizing objects (words) through the setting unit of the object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention (top) and the result of recognizing characters (bottom). FIGS. 5 and 6 illustrate an image recognition technology of an object recognition detection system for nuclear power plant control logic characters according to an embodiment of the present invention. FIG. 7 is a diagram illustrating the operation of an object recognition detection system for nuclear power plant control logic characters according to an embodiment of the present invention. FIG. 8 is an overall flowchart illustrating a method using an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention. Specific details for implementing the invention

[0013] The present invention will be described in detail below with reference to the contents described in the attached drawings. However, the present invention is not limited or restricted by exemplary embodiments. Identical reference numerals in each drawing indicate components that perform substantially the same function.

[0014] The purpose and effects of the present invention may be naturally understood or become clearer through the following description, and the purpose and effects of the present invention are not limited solely to the description below. Furthermore, in describing the present invention, if it is determined that a detailed description of known technology related to the present invention may unnecessarily obscure the essence of the present invention, such detailed description will be omitted.

[0015] FIG. 1 is a configuration diagram showing an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention.

[0016] As illustrated in FIG. 1, the object recognition detection system (10) for nuclear power plant control logic characters includes a collection unit (100), a setting unit (200), a detection unit (300), and a recognition unit (400).

[0017] The collection unit (100) is configured to collect data on the size and coordinate values ​​of each symbol in the nuclear power plant control logic drawing.

[0018] The setting unit (200) recognizes the character symbol as an individual object using lines within the drawing, performs data learning through image recognition of the character symbol as a single object, and sets the character symbol with noise as a single image based on the learned data.

[0019] The setting unit (200) according to the present embodiment recognizes a character symbol as a single image based on learned data and sets it as a word or character, wherein a character symbol containing noise based on learned data is set as an image in the word unit rather than in the character unit.

[0020] FIG. 2 is a diagram illustrating the recognition of characters as individual objects through a setting unit of an object recognition detection system for nuclear power plant control logic characters according to an embodiment of the present invention. As shown in FIG. 2, characters are recognized as individual objects, and based on learned data, character symbols are recognized as a single image to be recognized as a word or character.

[0021] FIG. 3 is a diagram showing individually recognized training data through a setting unit of an object recognition detection system for nuclear power plant control logic characters according to an embodiment of the present invention. As shown in FIG. 3, individually recognized training data is stored in the form of individual folders.

[0022] FIG. 4 shows the result of recognizing objects (words) through the setting unit of the object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention (top) and the result of recognizing characters (bottom). As shown in FIG. 4, based on the learned training data, character symbols containing noise are recognized by setting them as images of a single word unit rather than a character unit.

[0023] As shown below in Fig. 4, if character symbols containing noise are recognized on a character-by-character basis, false positives may occur.

[0024] FIGS. 5 and FIGS. 6 illustrate an image recognition technology for an object recognition detection system of a nuclear power plant control logic character according to an embodiment of the present invention. FIG. 5 shows a Tesseract result, and FIG. 6 shows a YOLOv3 result.

[0025] The setting unit according to the present embodiment learns characters with a lot of noise within the drawing as a single object (word) unit using image recognition technology (YOLOv3). Based on the learned data, recognition is performed at the word level rather than the character level. Character units other than word-level characters are recognized as a single image feature.

[0026] The detection unit (300) detects objects by determining the recognizable size range of each symbol and setting a grace period for the symbol range.

[0027] The detection unit (300) determines the recognizable size range using the smallest and largest values ​​for each symbol, and sets the recognition range according to the maximum and minimum values ​​for each symbol such that the minimum value of the symbol size < the recognition range < the maximum value, and detects objects by setting a grace period for the range to match the size that yields relatively high accuracy.

[0028] For example, in the case of AND, if the smallest symbol is 20 pixels and the largest is 50 pixels, the condition should be 20 < recognition range < 50, but the range is expanded by about 5% to 10%. Here, small symbols are made smaller and large symbols are made larger.

[0029] By giving a grace period equal to the range where high accuracy is obtained, recognition is performed within the range of 19 < recognition range < 53 to obtain high accuracy.

[0030] The recognition unit (400) is configured to immediately recognize symbols when recognizing drawings by setting a range value that yields high accuracy for each symbol. The recognition unit (400) sets a range value that yields high accuracy for each symbol through this data so that it can immediately recognize the nuclear power plant drawings when recognizing them in the future.

[0031] FIG. 7 is a diagram illustrating the operation of an object recognition detection system for nuclear power plant control logic characters according to an embodiment of the present invention.

[0032] As illustrated in FIG. 7, the recognition accuracy optimization system for adjusting the recognition range per symbol of a nuclear power plant control logic drawing according to one embodiment of the present invention divides the drawing while maintaining the ratio of the nuclear power plant drawing as much as possible before collecting data of the size and coordinate values ​​per symbol from the collection unit (100). At this time, for example, the nuclear power plant drawing is 4,678*3308 pixels.

[0033] Next, the collection unit (100) performs image division a total of 4 times by adding half (e.g., 150 pixels) of the largest symbol along the x-axis, y-axis, and xy-axis because there may be cut symbols, and obtains the size and coordinate values ​​of all symbols through duplicate recognition.

[0034] Next, the setting unit (200) recognizes the character symbol as an individual object using lines within the drawing, performs data learning through image recognition of the character symbol as a single object, and sets the character symbol with noise as a single image based on the learned data.

[0035] Next, the detection unit (300) sets the minimum / maximum value of each symbol as the recognition range. Symbols that do not fall within the appropriate size criteria for each symbol are excluded even if they are recognized symbols. For example, if a cut symbol has higher accuracy than an intact symbol, the cut symbol should be recognized by prioritizing higher accuracy; however, it is not recognized because the size is reduced by the amount of cutting and does not fall within the recognition range.

[0036] Next, the recognition unit (400) excludes symbols with high accuracy from among the recognized symbols that have the same coordinates and deletes them.

[0037] FIG. 8 is an overall flowchart illustrating a method using an object recognition detection system for nuclear power plant control logic characters according to one embodiment of the present invention.

[0038] First, the object recognition detection system for nuclear power plant control logic characters (hereinafter, the system) collects data on the size and coordinate values ​​of each symbol in the nuclear power plant control logic drawing (a).

[0039] Next, the system recognizes the character symbols as individual objects based on lines within the drawing, performs data learning through image recognition of the character symbols as a single object, and sets the character symbols with noise as a single image based on the learned data (b).

[0040] Next, the system detects objects by determining the recognizable size range of each symbol and setting a grace period for the symbol range (c).

[0041] And the system recognizes symbols when recognizing drawings by setting a range value that yields high accuracy for each symbol (d).

[0042] In step (a), the system divides the drawing while maintaining the aspect ratio of the nuclear power plant drawing as much as possible before collecting data of the size and coordinate values ​​per symbol through the collection unit.

[0043] Also, in step (a), the system performs duplicate image segmentation by adding half of the largest symbol to the x-axis, y-axis, and xy-axis to obtain the size and coordinate values ​​of all symbols.

[0044] (b) In step, the system recognizes character symbols as a single image based on data learned through the setting unit and sets them as a word or character, but character symbols containing noise based on the learned data are set as word-unit images rather than character-unit images.

[0045] (c) In step, the system determines the recognizable size range using the smallest and largest values ​​for each symbol through the detection unit, and sets the recognition range according to the maximum and minimum values ​​for each symbol such that the minimum symbol size < recognition range < maximum symbol size, and detects objects by setting a range deferral to the size that yields relatively high accuracy.

[0046] And in step (d), the system deletes the symbols among those recognized through the recognition unit, excluding the symbols with the highest accuracy among those with the same coordinates.

[0047] According to the present invention, when using an artificial intelligence-based control logic drawing verification tool to prevent false positives in drawing symbol recognition, a recognizable area for each symbol is set to prevent false positives, thereby improving the recognition rate. In addition, it reduces the false positive rate of reading symbols as characters. When recognizing characters as objects, it showed an average accuracy of 92.87%, which is an improvement of 27.54% compared to when using the character recognition tool Tesseract.

[0048] (Recognition Rate Comparison Table)

[0049]

[0050] According to the present invention, when using an AI-based control logic drawing verification tool to prevent false positives in drawing symbol recognition, a recognizable area for each symbol is set to prevent false positives, thereby improving the recognition rate. Additionally, it reduces the false positive rate of reading symbols as text.

Claims

Claim 1 A nuclear power plant control logic character object recognition and detection system comprising: a collection unit that collects data on the size and coordinate values ​​of each symbol in a nuclear power plant control logic drawing; a setting unit that recognizes character symbols as individual objects based on lines within the drawing, performs data learning through image recognition of the character symbols as a single object unit, and sets character symbols containing noise as a single image based on the learned data; a detection unit that detects objects by setting a recognizable size range for each symbol and setting a grace period for the symbol range; and a recognition unit that recognizes symbols during drawing recognition by setting a range value that yields high accuracy for each symbol, wherein the setting unit recognizes character symbols as a single image based on the learned data and sets them as words or characters, and character symbols containing noise based on the learned data are set as word-unit images rather than character-unit images. Claim 2 delete Claim 3 The object recognition detection system for nuclear power plant control logic characters according to claim 1, wherein the collection unit divides the drawing while maintaining the ratio of the nuclear power plant drawing as much as possible before collecting data of the size and coordinate values ​​per symbol. Claim 4 The object recognition detection system for nuclear power plant control logic characters according to claim 1, characterized in that the collection unit performs redundant image segmentation by adding half of the largest symbol along the x-axis, y-axis, and xy-axis to obtain the size and coordinate values ​​of all symbols. Claim 5 The object recognition detection system for nuclear power plant control logic characters according to claim 1, wherein the detection unit determines a recognizable size range using the smallest and largest values ​​for each symbol, and sets the recognition range according to the maximum and minimum values ​​for each symbol such that the minimum value of the symbol size < recognition range < maximum value, and detects objects by setting a grace period for the range to match the size that yields relatively high accuracy. Claim 6 A nuclear power plant control logic character object recognition detection system according to claim 1, wherein the recognition unit excludes symbols with high accuracy among the recognized symbols having the same coordinates and deletes the rest. Claim 7 A method for detecting object recognition of characters in nuclear power plant control logic using an object recognition detection system, comprising: (a) a step in which the system collects data on the size and coordinate values ​​of each symbol in a control logic drawing of a nuclear power plant; (b) a step in which the system recognizes character symbols as individual objects by lines within the drawing, performs data learning through image recognition of the character symbols as a single object unit, and sets the character symbols containing noise as a single image based on the learned data; (c) a step in which the system determines a recognizable size range for each symbol and sets a deferral for the symbol range to detect objects; and (d) a step in which the system sets a range value that yields high accuracy for each symbol and recognizes the symbol during drawing recognition, wherein the setting step includes a step of recognizing the character symbols as a single image based on the learned data and setting them as a word or character, wherein character symbols containing noise based on the learned data are set as word-unit images rather than character-unit images.

Images