Rain and fog weather instrument character recognition method and system integrating large model prior knowledge

Abstract

The present application provides a method and system for instrument character recognition in rainy and foggy weather that integrates large-scale model prior knowledge, wherein the method includes: when the marine instrument is in a rainy and foggy environment, by collecting the hydrophobic property data of the anti-corrosion coating surface and the dynamic parameters of the environment, using the large language model analysis to generate an optical compensation strategy, dynamically adjusting the wavelength of the light source, the polarization angle and the filter transmittance, and forming an interference fringe pattern that characterizes the characteristics of the droplets and salt crystals. The salt crystal scattering noise and the droplet refraction distortion are suppressed and corrected respectively through dual-branch processing, and the recognition results are optimized by combining spatial correlation technology with the character stroke continuity prior rule. The present application improves the anti-interference ability and accuracy of marine instrument character recognition in rainy and foggy environments.

Description

Technical Field

[0001] The present application relates to the field of computer vision and optical detection technology, and in particular to a method and system for character recognition of rain and fog weather instruments that integrates large-scale model prior knowledge. Background Art

[0002] Navigation instruments are susceptible to interference from droplet adhesion and salt crystal deposition in harsh environments such as salt spray and rain fog. This can cause refraction distortion and scattering noise on instrument characters, seriously affecting visual recognition accuracy. A method is needed that can dynamically adapt to environmental changes, suppress optical interference, and accurately recognize characters to ensure the safety and reliability of navigation operations.

[0003] One solution to this problem is a recognition method based on multispectral imaging and convolutional neural networks. It uses multi-band light sources to collect instrument images under different environmental conditions, uses neural networks to suppress noise and distortion in the images, and finally extracts instrument information through a character segmentation algorithm.

[0004] This approach relies on a fixed optical parameter configuration, resulting in residual scattering noise and insufficient correction of refractive distortion even in extreme rain and fog conditions. Furthermore, the neural network lacks prior knowledge of instrument character structure, resulting in poor recognition robustness when strokes are broken or connected. Summary of the Invention

[0005] The present application provides a method and system for recognizing characters of instruments in rainy and foggy weather by integrating large-scale model prior knowledge, so as to solve the problems of low anti-interference ability and poor accuracy of character recognition of navigation instruments in rainy and foggy environments in the prior art.

[0006] In a first aspect, the present application provides a method for character recognition of rain and fog weather instruments that integrates large model prior knowledge, comprising:

[0007] Obtain surface hydrophobicity data and environmental dynamic parameters of the anti-corrosion coating of marine instruments in rain and fog environments;

[0008] Performing correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy;

[0009] generating an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy;

[0010] Performing dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively;

[0011] The noise suppression map and the distortion correction map are spatially correlated to obtain a spatial correlation map, and the spatial correlation map is iteratively optimized in combination with the character stroke continuity prior rule in the preset large model prior knowledge base to generate an instrument character recognition result.

[0012] Optionally, the iterative optimization of the spatial association graph in combination with the character stroke continuity prior rule in the preset large model prior knowledge base to generate the instrument character recognition result includes:

[0013] Extracting candidate line segments related to the stroke directions of instrument characters from the spatial association map;

[0014] Matching the candidate line segments with the character stroke continuity priori rule to generate a candidate line segment set;

[0015] Iteratively optimizing the spatial correlation map according to the set of candidate line segments to generate an optimized spatial correlation map;

[0016] Candidate character boundaries that match the predefined character structure template in the character stroke continuity prior rule are extracted from the optimized spatial association map to generate an instrument character recognition result.

[0017] Optionally, matching the candidate line segments with the character stroke continuity prior rule to generate a candidate line segment set includes:

[0018] Dividing a set of line segment direction types according to the extension direction and length ratio of the candidate line segments in the spatial association map;

[0019] Extracting a stroke turning angle range corresponding to the line segment direction type set from the character stroke continuity prior rule;

[0020] Grouping the candidate line segments into candidate line segment clusters according to corresponding extension directions;

[0021] Comparing the distance between the start and end points of adjacent line segment pairs in the candidate line segment cluster with a preset distance tolerance, where the start and end point distance is the distance between the start end point of one candidate line segment and the end end point of the other candidate line segment in the adjacent line segment pair;

[0022] For each pair of adjacent line segments whose distance between the first and last endpoints is less than the distance tolerance, calculating the angle fluctuation amplitude of the adjacent line segment pair based on the stroke turning angle range;

[0023] Based on the angle fluctuation amplitude, a candidate line segment set is generated.

[0024] Optionally, generating a set of candidate line segments based on the angle fluctuation amplitude includes:

[0025] Extracting a curvature change allowable threshold corresponding to the line segment direction type set from the character stroke continuity prior rule;

[0026] Eliminating adjacent line segment pairs whose angle fluctuation amplitude exceeds the curvature change allowable threshold from the candidate line segment cluster to generate a target line segment cluster;

[0027] Taking the center point of the target line segment cluster as a reference, extending a virtual connecting line along the direction of curvature change, and determining whether the virtual connecting line is consistent with the stroke extension trend of the character structure in the character stroke continuity prior rule;

[0028] When it is determined to be consistent, dynamically adjusting the angle threshold of adjacent line segments in the target line segment cluster;

[0029] Adjacent line segment pairs that meet the dynamically adjusted angle threshold are connected and integrated to generate a set of candidate line segments.

[0030] Optionally, iteratively optimizing the spatial correlation map based on the candidate line segment set to generate an optimized spatial correlation map includes:

[0031] According to the spatial distribution relationship between the endpoints of the line segments in the candidate line segment set, a pixel confidence enhancement region associated with the character stroke overlap probability is divided in the spatial association map;

[0032] Based on the stroke connection topology constraint relationship in the character stroke continuity prior rule, bidirectional weighted diffusion is performed on the pixel confidence enhancement area to construct a confidence distribution network covering the complete character stroke outline;

[0033] According to the confidence distribution network and the set of candidate line segments, the confidence levels of the pixels in the spatial association map are dynamically enhanced or attenuated to generate an optimized spatial association map.

[0034] Optionally, the performing correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy includes:

[0035] Binding the coating contact angle change rate in the surface hydrophobicity characteristic data and the salt spray concentration gradient in the environmental dynamic parameters in a time series manner to generate a coupling characteristic vector of the environment and the material;

[0036] Retrieving matching historical optical compensation cases in the large language model based on the coupling feature vector, and extracting a set of cases whose similarity with the current coupling feature vector reaches a preset similarity threshold from the historical optical compensation cases;

[0037] Based on the case set, calculate the predicted value of droplet distribution density and salt crystal deposition thickness on the surface of the anti-corrosion coating of the marine instrument under the current environmental conditions;

[0038] generating corresponding light source wavelength adjustment suggestion information and polarization angle adjustment suggestion information according to the droplet distribution density prediction value;

[0039] generating corresponding filter transmittance curve adjustment suggestion information according to the salt crystal deposition thickness prediction value;

[0040] An optical compensation strategy is generated by integrating the light source wavelength adjustment suggestion information, the polarization angle adjustment suggestion information, and the filter transmittance curve adjustment suggestion information.

[0041] Optionally, generating an interference fringe pattern corresponding to the surface of the anti-corrosion coating of a marine instrument according to the optical compensation strategy includes:

[0042] Based on the light source wavelength adjustment suggestion information in the optical compensation strategy, the wavelength of the incident light beam of the illumination light source is adjusted so that the incident light beam excites a scattered characteristic light beam representing the morphology of the droplets on the surface of the anti-corrosion coating;

[0043] Based on the polarization angle adjustment suggestion information in the optical compensation strategy, the polarization angle of the circularly polarized light source is adjusted to suppress stray reflected light caused by salt crystal deposition on the surface of the anti-corrosion coating of the marine instrument;

[0044] Adjusting the filter transmittance curve of the filter array based on the filter transmittance curve adjustment suggestion information in the optical compensation strategy to enhance the interference contrast between the scattered characteristic light beam and the refracted characteristic light beam;

[0045] An interference light intensity distribution is formed by the interaction between the scattered characteristic light beam and the refracted characteristic light beam in the imaging plane, and an interference fringe pattern is generated according to the interference light intensity distribution.

[0046] In a second aspect, the present application provides a rain and fog weather instrument character recognition system that integrates large model prior knowledge, including:

[0047] An acquisition module is used to obtain surface hydrophobicity characteristic data and environmental dynamic parameters of the anti-corrosion coating of the navigation instrument when the navigation instrument is in a rainy and foggy environment;

[0048] A correlation analysis module, configured to perform correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy;

[0049] A first generating module is used to generate an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy;

[0050] A second generating module is used to perform dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively;

[0051] The spatial association module is used to spatially associate the noise suppression map with the distortion correction map to obtain a spatial association map, and iteratively optimize the spatial association map in combination with the character stroke continuity prior rules in the preset large model prior knowledge base to generate instrument character recognition results.

[0052] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described in the first aspect for character recognition of instruments in rainy and foggy weather that integrates large model prior knowledge.

[0053] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, which, when executed by a processor, implements a method for character recognition of rain and fog weather instruments that integrates large-scale model prior knowledge as described in any one of the first aspects.

[0054] In the present application, a method for recognizing instrument characters in rainy and foggy weather that integrates large-scale model prior knowledge is provided, the method comprising: obtaining surface hydrophobic property data and environmental dynamic parameters of the anti-corrosion coating of the marine instrument when the marine instrument is in a rainy and foggy environment; performing correlation analysis on the surface hydrophobic property data and the environmental dynamic parameters through a pre-trained large language model to generate an optical compensation strategy; generating an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy; performing dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern, respectively; performing spatial correlation on the noise suppression pattern and the distortion correction pattern to obtain a spatial correlation pattern, and iteratively optimizing the spatial correlation pattern in combination with the character stroke continuity prior rule in a preset large-scale model prior knowledge base to generate an instrument character recognition result.

[0055] The technical solution provided by this application has the following beneficial effects:

[0056] This application uses real-time sensing of instrument surface conditions and environmental changes to provide a data foundation for optical compensation. It dynamically adapts to varying rain and fog conditions and optimizes light source parameter configuration. It simultaneously captures salt crystal scattering and droplet refraction characteristics to create an interference-resistant optical representation. It also eliminates particle noise and geometric deformation separately. It also fuses multimodal features and utilizes prior knowledge to restore stroke structure.

[0057] Furthermore, the present application also extracts candidate line segments from the spatial association map, screens the valid line segment set in combination with the character stroke continuity rule, matches the predefined character template after iterative optimization, and finally outputs the recognition result.

[0058] In addition, it effectively solves the problem of character stroke breakage / adhesion in rainy and foggy environments, and improves the accuracy of character structure restoration in complex interference scenes through an optimization process guided by prior knowledge.

[0059] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0061] Figure 1 A flowchart of a method for character recognition of rain and fog weather instruments that integrates large model prior knowledge provided in an embodiment of the present application;

[0062] Figure 2 A schematic diagram of the structure of a rain and fog weather instrument character recognition system that integrates large-scale model prior knowledge provided in an embodiment of the present application;

[0063] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0065] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0066] Researchers have found that traditional visual recognition methods of marine instruments are seriously interfered with by salt crystal deposition and droplet adhesion in rainy and foggy environments. Existing technologies have difficulty dynamically adapting to environmental changes and effectively restoring damaged character structures. Based on this, an embodiment of the present application provides a method for instrument character recognition in rainy and foggy weather that integrates large model prior knowledge. This method perceives the hydrophobic properties of the instrument surface and environmental parameters in real time, uses a large language model to dynamically generate an optical compensation strategy, combines dual-branch processing to separate noise and distortion features, and finally achieves anti-interference recognition through iterative optimization guided by prior knowledge. The technical solution of this application can be applied to automatic instrument recognition scenarios in severe weather conditions such as navigation and aviation.

[0067] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0068] Figure 1 A flowchart of a method for character recognition of rain and fog weather instruments that integrates large model prior knowledge is provided in an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0069] Step 101: When the marine instrument is in a rainy and foggy environment, surface hydrophobic property data and environmental dynamic parameters of the marine instrument's anti-corrosion coating are obtained.

[0070] In this step, surface hydrophobicity data reflects the anti-corrosion coating's ability to repel droplets and is used to predict rain and fog adhesion patterns. Environmental dynamic parameters represent real-time monitoring of key external factors affecting optical imaging quality, such as salt spray concentration, humidity, and temperature.

[0071] In this embodiment, a contact angle meter collects contact angle data from a droplet on an instrument's anti-corrosion coating. Environmental sensors simultaneously acquire salt spray concentration and humidity data. These two data types are aligned by time stamp and then input into the system. The contact angle data reflects the coating's current hydrophobicity, while the environmental parameters are used to assess interference intensity.

[0072] For example, taking a marine engine instrument displaying "speed 1500RPM" as an example, the coating contact angle was measured to be 120 degrees (indicating high hydrophobicity), and at the same time, a high salt spray concentration and saturated humidity were detected. These data indicate that severe droplet adhesion and salt crystal scattering will occur.

[0073] Step 102: performing correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy.

[0074] In this step, the optical compensation strategy includes adjustment schemes such as light source wavelength, polarization angle, and filter parameters to offset environmental interference.

[0075] In this application's example, a large language model analyzes the correlation between contact angle data and salt spray concentration. When the contact angle exceeds a threshold and the salt spray concentration is high, a short-wavelength light source is output to enhance the contrast at the droplet edge. The polarizer angle is adjusted based on the humidity value to suppress reflections from salt crystals. The model generates parameter combinations through historical case matching.

[0076] For example, for the aforementioned 120-degree contact angle and high salt fog environment, the model selects 450nm blue light (short wavelength) to highlight the droplet outline and sets the polarizer at a 60-degree angle to filter out salt crystal stray light, forming a targeted optical compensation solution.

[0077] Step 103: Generate an interference fringe pattern corresponding to the surface of the marine instrument anti-corrosion coating according to the optical compensation strategy.

[0078] In this step, the interference fringe pattern represents a two-dimensional grayscale image formed by the interference of scattered light (salt crystals) and refracted light (droplets), which contains intensity and phase information.

[0079] In this embodiment, a multispectral light source and polarizer are configured according to a compensation strategy. After illuminating the instrument surface, the reflected light passes through a filter array to separate the characteristic wavelengths. An imaging sensor collects the interfering light intensity distribution, which is then scattered by salt crystals to form radial fringes, and refracted by liquid droplets to form concentric fringes, creating a composite image.

[0080] For example, a "1500 RPM" meter is illuminated with 450nm blue light and imaged after polarization filtering. The numeral "1" is formed by the droplet, creating parallel stripes with a spacing of 2mm (determined by the wavelength and the curvature of the droplet), while the arc of the letter "R" exhibits curved stripes.

[0081] Step 104: performing dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively.

[0082] In this step, the noise suppression map represents the image after removing the salt crystal scattering noise, preserving the main body of the characters. The distortion correction map represents the image after correcting the geometric deformation caused by the droplet refraction.

[0083] In this embodiment, the first branch uses a polarization difference algorithm to suppress salt crystal noise by comparing image differences in different polarization directions. The second branch restores distorted strokes to their original shapes based on reverse calculations of the refracted light path. Both branches process the same interference pattern in parallel.

[0084] For example, in the "1500RPM" processing, the first branch removes the salt crystal noise around the letter "P", and the second branch corrects the bending deformation of the number "5" caused by the droplets and restores its standard semicircular structure.

[0085] Step 105: spatially correlate the noise suppression map and the distortion correction map to obtain a spatial correlation map, and iteratively optimize the spatial correlation map in combination with the character stroke continuity prior rule in the preset large model prior knowledge base to generate an instrument character recognition result.

[0086] In this step, the spatial correlation map represents a three-dimensional data volume that fuses the features of the two types of maps, including coordinates, intensity, and confidence information. The character stroke continuity prior rules represent the stroke connection probability library provided by the large model. For example, the number "1" should be a continuous straight line. Instrument character recognition results are the final readable character information output after environmental interference suppression, feature fusion, and prior knowledge optimization. Specifically, they accurately reproduce the actual content displayed by marine instruments (such as "speed 1500RPM") in rainy and foggy environments. The stroke continuity and geometric structure of each character are consistent with the actual instrument display, eliminating artifacts caused by salt crystal scattering and deformation caused by droplet refraction. This ensures the integrity and legibility of numbers, letters, and unit symbols, meeting the requirements for accurate interpretation of instrument data in marine operations.

[0087] In this embodiment, the two atlases are aligned by pixel coordinates, and confidence weights are calculated for the overlapping regions. Prior rules guide optimization: when a stroke interruption in the digit "1" is detected, the missing pixels are filled in based on line continuity; and connected strokes are segmented according to the character template.

[0088] For example, the "P" in "RPM" is interrupted by noise, and the prior rule recognizes that it should contain a closed ring structure, and automatically connects the broken edges; the middle of the number "0" is blurred by droplets, and the edge confidence is enhanced according to the circular template.

[0089] This method dynamically senses the environment and coating status to generate targeted optical compensation solutions, uses interferometric imaging to separate the interference features of salt crystals and droplets, suppresses noise and corrects deformation through dual-branch processing, and finally achieves accurate recognition by combining prior knowledge of character structure.

[0090] In order to solve the problem of stroke breakage and deformation in instrument character recognition in rainy and foggy environments and further improve recognition accuracy, in some embodiments, step 105: iteratively optimizing the spatial association map in combination with the character stroke continuity prior rules in the preset large model prior knowledge base to generate instrument character recognition results includes:

[0091] Step 201: extracting candidate line segments related to the stroke directions of instrument characters from the spatial association map.

[0092] In step 201, the candidate line segments refer to a set of line segments of suspected character strokes extracted from the spatial association map, each line segment contains starting point coordinates, end point coordinates and direction angle information, which are used to preliminarily represent the character structure.

[0093] In an embodiment of the present application, connected areas in a spatial correlation map are extracted using an edge detection algorithm, and a set of line segments with consistent directions is generated after calculating the main axis directions of each area, and interfering line segments with too short lengths or disordered directions are filtered out.

[0094] Step 202: Match the candidate line segments with the character stroke continuity priori rule to generate a candidate line segment set.

[0095] In step 202, the candidate segment set refers to the combination of segments retained after screening based on a priori character stroke continuity rules. These segments conform to the stroke characteristics of the predefined character structure in terms of directional continuity, turning angles, and curvature changes. Specifically, they include two types of valid segments: one is the original complete character stroke segments in the spatial association graph, and the other is the broken stroke segments completed using a priori rules. This set serves as the basic input for subsequent iterative optimization to ensure the structural integrity and geometric accuracy of the final recognized character. For example, when recognizing the oil pressure gauge "25.5MPa", the set includes the complete S-shaped stroke segment of the number "5" after rule completion and the straight stroke segments of the letter "MPa".

[0096] In an embodiment of the present application, after the candidate line segments are grouped by direction, they are matched with the character templates in the rule library to screen out line segment combinations with turning angles within a preset range and smooth curvature changes to form a candidate line segment set.

[0097] Step 203: performing iterative optimization on the spatial correlation map according to the candidate line segment set to generate an optimized spatial correlation map.

[0098] In step 203, the optimized spatial correlation map refers to a feature map after correction of stroke continuity, and its pixel confidence distribution is more consistent with the real character structure.

[0099] In an embodiment of the present application, based on the connection relationship between the line segments in the candidate line segment set, the pixel confidence of the line segment break area in the original map is enhanced, and the pixel confidence of the line segment intersection conflict area is attenuated, and the optimized map is output after multiple iterations.

[0100] Step 204: extracting candidate character boundaries that match the predefined character structure template in the character stroke continuity prior rule from the optimized spatial association map to generate an instrument character recognition result.

[0101] In step 204, the candidate character boundaries refer to closed contour lines in the optimized graph that meet the geometric features of the character template and are used for final recognition.

[0102] In an embodiment of the present application, connected domain boundaries are extracted on the optimized graph, shape matching is performed with character templates in the rule base, and boundaries with similarity reaching a threshold are screened out as recognition results.

[0103] Here's a specific example:

[0104] For example, a marine hydraulic gauge displaying "Oil Pressure 25.5 MPa" was used in an environment with a measured coating contact angle of 115 degrees and high salt spray concentration. Interference fringes were generated using 470nm blue-green light (calculated using an optical compensation strategy based on the relationship between droplet curvature and wavelength-refractive index). The numeral "5" formed wavy fringes with a spacing of 1.8mm (spacing = wavelength × droplet curvature radius / refractive index difference) due to refraction from the droplets. The diagonal lines of the letter "M" appeared as intermittent bright spots due to scattering from salt crystals. In a two-branch processing, the first branch eliminated salt crystal noise to the right of the "M," while the second branch corrected the S-shaped distortion at the bottom of the numeral "5" caused by the droplets. During spatial correlation optimization, a priori rules identified the doubly curved line segments characteristic of the numeral "5," complemented the broken line segments at the troughs based on curvature continuity, and restored the circular structure of the blurred "a" in the unit "MPa" using a standard font template. The final output was a complete and accurate recognition result of "25.5 MPa." The 1.8mm fringe spacing is calculated using an optical formula: the selected wavelength of 470nm multiplied by the average droplet curvature radius of 3.8mm (converted by the contact angle) divided by the oil-water refractive index difference of 1.33.

[0105] In the embodiment of the present application, the above steps are used to achieve structural repair and accurate recognition of broken and deformed characters under rain and fog interference, thereby improving the reliability of instrument readings in harsh environments.

[0106] To address the matching difficulties caused by broken and deformed instrument character strokes in rainy and foggy environments and to further improve the accuracy of character segment screening, in some embodiments, step 202: matching the candidate segments with the character stroke continuity prior rules to generate a set of candidate segments includes:

[0107] Step 301: Divide the line segment direction type set according to the extension direction and length ratio of the candidate line segment in the spatial association map.

[0108] In step 301, the line segment direction type set refers to the categories into which the candidate line segments are divided according to their main extension directions in the spatial association map, including four basic types: horizontal, vertical, left-slanting and right-slanting. Each type of line segment has similar length ratio characteristics.

[0109] In an embodiment of the present application, the candidate line segments are classified into a preset direction type set by calculating the angle between each line segment and the coordinate axis and combining the ratio of the line segment length to the average stroke length of the character.

[0110] Step 302: extracting a stroke turning angle range corresponding to the line segment direction type set from the character stroke continuity prior rule.

[0111] In step 302, the stroke turning angle range refers to the maximum angle change threshold allowed when the character stroke is converted into different direction types.

[0112] In an embodiment of the present application, the type of character currently being processed (such as a number or a letter) is queried from a priori rule library, and a reasonable turning angle range for the character when converting the type in each direction is extracted.

[0113] Step 303: Group the candidate line segments into candidate line segment clusters according to corresponding extension directions.

[0114] In step 303, the candidate line segment cluster refers to a group of line segments of the same direction type and adjacent in spatial position, which is used for local stroke continuity analysis.

[0115] In an embodiment of the present application, line segments with the same direction type are clustered according to their coordinate positions in the atlas to form a combination of line segments that may belong to the same stroke in space.

[0116] Step 304: Compare the head-end and tail-end distances of adjacent line segment pairs in the candidate line segment cluster with a preset distance tolerance, where the head-end and tail-end distances are the distances between the head end point of one candidate line segment and the tail end point of the other candidate line segment in the adjacent line segment pair.

[0117] In step 304, the spacing tolerance is the maximum endpoint distance threshold used to determine whether two line segments are likely to belong to the same stroke, and is dynamically adjusted based on the character size. Adjacent line segments, in the spatial association graph, are two or more candidate line segments that extend in similar directions and have a spacing between their endpoints that is less than a preset tolerance.

[0118] In an embodiment of the present application, the actual distance between the endpoints of adjacent line segments is calculated and compared with a preset tolerance value to screen out possible connected line segment pairs.

[0119] Step 305: For each pair of adjacent line segments whose distance between the first and last endpoints is less than the distance tolerance, calculate the angle fluctuation amplitude of the adjacent line segment pair based on the stroke turning angle range.

[0120] In step 305, the angle fluctuation amplitude refers to the degree of deviation between the angle difference between the direction changes of adjacent line segments and the standard turning angle in the prior rule.

[0121] In an embodiment of the present application, for the line segment pairs that pass the spacing screening, the angle of their extension direction is calculated and compared with the standard turning angle range in the rule library to obtain the fluctuation amplitude.

[0122] Step 306: Generate a set of candidate line segments based on the angle fluctuation amplitude.

[0123] In the embodiment of the present application, line segment pairs whose fluctuation amplitude exceeds a threshold are eliminated, and line segments that meet the conditions are merged according to spatial relationships to form a final candidate set.

[0124] Here's a specific example:

[0125] Taking the identification of the digit "5" in the oil pressure gauge reading "25.5MPa" as an example, three broken line segments were extracted from the spatial correlation map: the first segment is 45 degrees right-slanting and 3.2 mm long (calculated from the 1.8 mm interference fringe spacing: 1.8 mm × 1.8 ≈ 3.2 mm); the second segment is horizontal and 2.1 mm long; and the third segment is 135 degrees left-slanting and 3.0 mm long. Based on the a priori rule of character stroke continuity, the standard turning angles for the digit "5" should be: right-slanting to horizontal 120 degrees (±15 degrees), horizontal to left-slanting 135 degrees (±10 degrees). After grouping the three segments by direction, the measured distances between adjacent segment endpoints were 0.5 mm and 0.6 mm, respectively (less than the preset 1 mm tolerance). The calculated angle between the first and second segments is 135 degrees (within the allowable range), and the angle between the second and third segments is 135 degrees (which meets the standard). Finally, the three segments are merged into a complete S-shaped stroke. The 1mm spacing tolerance is set based on the 20% empirical value of a digit height of 5mm to ensure effective connection of broken strokes.

[0126] In the embodiment of the present application, the above steps achieve accurate screening and reorganization of broken strokes, effectively solving the recognition error problem caused by broken character strokes in rainy and foggy environments, and improving the ability to restore character structure in complex interference scenarios.

[0127] To address the matching difficulties caused by broken and deformed instrument character strokes in rainy and foggy environments and to further improve the accuracy of character segment screening, in some embodiments, step 306: generating a set of candidate line segments based on the angle fluctuation amplitude includes:

[0128] Step 401: extracting a curvature change allowable threshold corresponding to the line segment direction type set from the character stroke continuity prior rule.

[0129] In step 401, the curvature change threshold refers to the maximum curvature change allowed when the character stroke is converted in different directions.

[0130] In an embodiment of the present application, based on the currently processed character type and line segment direction type, a corresponding curvature change threshold range is extracted from the prior rule library to determine the rationality of the line segment turning.

[0131] Step 402: Eliminate adjacent line segment pairs whose angle fluctuation amplitude exceeds the curvature change allowable threshold within the candidate line segment cluster to generate a target line segment cluster.

[0132] In step 402 , the target line segment cluster refers to the line segment grouping retained after screening by the curvature change threshold, wherein the turning angles and curvature changes of adjacent line segment pairs conform to the character structure characteristics.

[0133] In an embodiment of the present application, the curvature variation of each adjacent line segment pair in the candidate line segment cluster is calculated, line segment pairs exceeding an allowable threshold are eliminated, and line segment combinations with smooth transitions are retained.

[0134] Step 403: Taking the center point of the target line segment cluster as a reference, extend a virtual connecting line along the curvature change direction, and determine whether the virtual connecting line is consistent with the stroke extension trend of the character structure in the character stroke continuity prior rule.

[0135] In step 403, the direction of curvature change is calculated from the geometric features of the candidate line segment cluster, and is specifically determined by the trend of change in the tangent direction of adjacent line segments in the line segment cluster. Relationship with the curvature change allowable threshold: the actual fluctuation amplitude of the curvature change direction must be less than or equal to the curvature change allowable threshold, and the two together constitute a dynamic constraint condition for judging the continuity of the stroke. The virtual connecting line refers to a hypothetical connecting line extending from the center point of the target line segment cluster as the starting point according to the curvature change direction of the line segment cluster, which is used to simulate the extension path of the ideal character stroke. The generation process is: first calculate the average curvature direction of each line segment in the target line segment cluster, and then extend a virtual baseline along this direction.

[0136] In an embodiment of the present application, the average curvature direction of each line segment in the target line segment cluster is calculated, and a virtual baseline is extended along the direction to match the standard character stroke extension trend in the rule library.

[0137] Step 404: When it is determined to be consistent, dynamically adjust the angle threshold of adjacent line segments in the target line segment cluster.

[0138] In step 404, dynamically adjusting the angle threshold refers to adaptively loosening or tightening the allowable range of the angles between adjacent line segments according to the degree of matching between the virtual connecting line and the standard stroke trend.

[0139] In the embodiment of the present application, when the virtual connecting line is highly consistent with the standard trend, the angle threshold is appropriately relaxed to allow for recoverable broken strokes; when the matching degree is low, the threshold is tightened to avoid incorrect connections.

[0140] Step 405: Connect and integrate adjacent line segment pairs that meet the dynamically adjusted angle threshold to generate a candidate line segment set.

[0141] In step 405, connection integration involves smoothly connecting adjacent line segment pairs that meet the adjusted angle threshold to form complete stroke segments. The target segment cluster is the grouping of segments that has undergone preliminary screening (eliminating pairs that do not meet the initial threshold), while the candidate segment set is the combination of segments within the target segment cluster that are ultimately confirmed as valid character strokes after virtual connecting line verification and dynamic threshold adjustment. The candidate segment set is a subset of the target segment cluster and has a higher confidence level for stroke continuity.

[0142] In this embodiment, for adjacent line segment pairs that pass the screening, smooth transitions are performed at their endpoints to generate continuous stroke segments, which are then added to the candidate line segment set. Adjacent line segment pairs that do not meet the dynamically adjusted angle threshold remain separated and will not be connected or integrated, but will remain in the target line segment cluster as candidates for subsequent processing.

[0143] Here's a specific example:

[0144] Taking the identification of the digit "5" in the oil pressure gauge reading "25.5 MPa" as an example, during the target segment cluster processing, the curvature variation threshold for the digit "5" was first extracted from a priori rules, which stipulates a maximum curvature variation of 40 degrees per millimeter (derived from statistics in a standard font library). The actual curvature variation of the first right-sloping segment and the second horizontal segment was measured to be 35 degrees per millimeter (calculated by dividing the segment angle of 135 degrees by the length of the turning point, 3.8 mm). The curvature variation of the second and third segments was 38 degrees per millimeter, both meeting the threshold requirements. A virtual S-shaped connecting line was extended from the center point of the three segments. Its curvature variation trend matched the standard template for the digit "5" in the rule library with a 92% match (calculated using a curvature similarity algorithm). Therefore, the threshold for the adjacent segment angle was dynamically adjusted from ±15 degrees to ±18 degrees. Finally, the three line segments are smoothly connected at the turning point. The connection between the first and second segments uses a transition arc with a radius of 2.5 mm (radius = line segment length × curvature compensation coefficient 0.8) to generate a complete and coherent candidate stroke for the number "5".

[0145] In the embodiment of the present application, the above steps realize intelligent screening and precise repair of broken strokes, effectively solving the recognition difficulty problem caused by broken and deformed character strokes in rainy and foggy environments, and improving the character recognition accuracy in complex interference scenarios.

[0146] To address the recognition difficulties caused by broken and deformed instrument character strokes in rainy and foggy environments and to further improve the accuracy and robustness of character recognition, in some embodiments, step 203: iteratively optimizing the spatial correlation map based on the candidate line segment set to generate an optimized spatial correlation map includes:

[0147] Step 501: Based on the spatial distribution relationship between the endpoints of the line segments in the candidate line segment set, a pixel confidence enhancement area associated with the character stroke overlap probability is divided in the spatial association map.

[0148] In step 501, the spatial distribution relationship between the endpoints of the line segments refers to the relative position and connection trend of the endpoints of each line segment in the candidate line segment set, including features such as endpoint spacing and extension direction consistency. This relationship is obtained by calculating the actual distance and direction angle between adjacent line segment endpoints, and is used to determine whether the broken line segments may belong to the same stroke. For example, in the recognition of the number "5", the endpoint spacing and direction change trend of the three line segments conform to the distribution law of S-shaped strokes. Pixel confidence enhancement areas refer to pixel areas that overlap or are adjacent to the spatial position of the line segments in the candidate line segment set in the spatial association map. These areas have a higher probability of the existence of character strokes.

[0149] In an embodiment of the present application, by calculating the spatial distance and directional consistency between the endpoints of the candidate line segments, pixel areas that may belong to the same character stroke are marked in the spatial association map to form a confidence enhancement area.

[0150] Step 502: Based on the stroke connection topology constraint relationship in the character stroke continuity prior rule, bidirectional weighted diffusion is performed on the pixel confidence enhancement area to construct a confidence distribution network covering the complete character stroke outline.

[0151] In step 502, the stroke connection topology constraint refers to the geometric connection rules of character strokes within the standard structure, including topological features such as stroke turning angles and curvature trends. This constraint is extracted from the large model prior knowledge base and is used to guide the propagation of confidence scores within the network. For example, the constraint for the digit "5" requires that the strokes have continuous S-shaped turns and smooth curvature changes. The character stroke overlap probability refers to the probability that a character stroke (rather than noise or distortion) actually exists in a certain area of ​​the spatial association map. The instrument character stroke direction refers to the directional characteristics of the strokes of nautical instrument characters (such as geometric trends such as straight lines and arcs). The complete character stroke outline refers to the closed stroke boundary without breaks or distortion after optimization. The stroke direction is used to initially screen candidate line segments; the overlap probability determines which line segment regions require enhancement; and the complete outline is the final result of the coordinated optimization of the first two. These three form a progressive correction chain from local features to global structure. The confidence distribution network is a pixel confidence propagation network established based on the stroke connection topology constraint relationship and is used to represent the complete outline of the character strokes.

[0152] In an embodiment of the present application, based on the connection relationship of the character strokes in the prior rules, starting from the central pixel of the confidence enhancement area, bidirectional weighted diffusion is performed along the stroke extension direction to gradually construct a confidence distribution network covering the entire character outline.

[0153] Step 503: Dynamically enhance or attenuate the confidence levels of the pixels in the spatial correlation map based on the confidence distribution network and the candidate line segment set to generate an optimized spatial correlation map.

[0154] In step 503, dynamic enhancement or attenuation refers to adjusting the confidence level of the pixel in the spatial association map based on the spatial relationship between the confidence distribution network and the candidate line segment set. Specifically, confidence enhancement is performed when the pixel in the spatial association map meets both of the following conditions: (1) it is located on a high-weight path in the confidence distribution network, and (2) it matches the geometric features of the line segments in the candidate line segment set. Conversely, confidence attenuation is performed when the pixel meets either of the following conditions: (1) it is located in a low-weight area in the confidence distribution network, and (2) it has a spatial position conflict with the candidate line segment set.

[0155] In an embodiment of the present application, for pixel points on high-weight paths in the confidence distribution network, which match the spatial position of the line segments in the candidate line segment set, their confidence is enhanced; for pixel points in low-weight areas or that conflict with the candidate line segments, their confidence is attenuated, and finally an optimized spatial correlation map is generated.

[0156] Here's a specific example:

[0157] Taking the identification of the digit "5" in the oil pressure gauge "25.5MPa" as an example, an S-shaped stroke region was delineated in the spatial correlation map based on the endpoint spacing (0.5mm and 0.6mm, respectively) and extension direction (45 degrees right, 0 degrees horizontal, and then 135 degrees left) of the three candidate line segments. This region's width was set to 1.2mm (determined by 40% of the average segment length of 3.1mm). Based on the double-bend topological characteristics required by the prior rule for the digit "5," a bidirectional diffusion process was performed from the center of the region toward both ends with a weight coefficient of 0.15 per pixel (calculated from the ratio of stroke width to length), constructing a complete S-shaped confidence network. The confidence of pixels at the two turning points (2.5mm radius of curvature) in the network was enhanced by 1.5 times, while the confidence of pixels in the central region (approximately 0.8mm width) blurred by droplet interference was reduced to 0.7 times, ultimately generating a clear and coherent optimized map of the digit "5." The weight coefficient 0.15 = line segment width 1.2 mm / (total line segment length 8.3 mm × 10), and the 1.5-fold enhancement coefficient is calculated based on the matching degree between the curvature of the turning point and the standard value.

[0158] In the embodiment of the present application, the above steps are used to achieve accurate repair and optimization of broken and deformed strokes, effectively solve the interference problem in character recognition in rainy and foggy environments, and improve the accuracy and reliability of instrument character recognition.

[0159] In order to solve the adaptability problem of the optical compensation strategy in rainy and foggy environments and further improve the accuracy of instrument character recognition, in some embodiments, step 102: performing correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy includes:

[0160] Step 601: Bind the coating contact angle change rate in the surface hydrophobicity characteristic data and the salt spray concentration gradient in the environmental dynamic parameters in a time series manner to generate a coupling feature vector of the environment and the material.

[0161] In step 601, the coating contact angle change rate refers to the change in the surface contact angle of the anti-corrosion coating per unit time, reflecting the dynamic trend of hydrophobicity. This data is obtained by using a contact angle meter to collect contact angle values ​​at multiple locations on the instrument surface in real time and calculating the slope of the change over time. This data is used to quantify the impact of rain and fog adhesion on the coating's hydrophobicity. For example, in the oil pressure gauge scenario, the measured contact angle change rate from an initial 120 degrees to 115 degrees is 5 degrees per hour. The salt spray concentration gradient refers to the rate of change of salt spray particle concentration per unit area, indicating the uneven distribution of salt spray deposition. This data is collected by multiple salt spray sensors placed around the instrument and the spatial differences in the detection values ​​of adjacent sensors are calculated. For example, in a marine scenario, the concentration above the instrument is measured to be 15% / cm higher than the side, forming a vertical concentration gradient. The coupled characteristic vector of the environment and material is a multidimensional feature representation formed by temporally and spatially correlating the surface hydrophobicity with environmental parameters. It is used to characterize the physical state of the material surface under specific environmental conditions.

[0162] In an embodiment of the present application, the rate of change of the coating contact angle over time is synchronously aligned with the salt spray concentration change curve, and a multidimensional vector containing a time series relationship is generated through a feature fusion algorithm.

[0163] Step 602: searching for matching historical optical compensation cases in the large language model based on the coupling feature vector, and extracting a set of cases whose similarity with the current coupling feature vector reaches a preset similarity threshold from the historical optical compensation cases.

[0164] In step 602, historical optical compensation cases refer to the set of optimal matching solutions for environment, material, and optical parameters learned by the large language model during training. The current coupled feature vector is a feature vector generated by real-time binding of currently collected surface hydrophobicity data (e.g., coating contact angle change rate) with currently detected environmental dynamic parameters (e.g., salt spray concentration gradient). Regarding its relationship to the coupled feature vector, the coupled feature vector is a general concept, while the current coupled feature vector specifically refers to an instantiated vector generated in real time during this processing. The case set is a group of successful optical compensation cases retrieved from the large language model's historical database that match the current environment-material characteristics. Each case contains a correspondence between specific environmental parameters, material properties, and the optimal optical compensation solution. This set is constructed by calculating the similarity score between the current coupled feature vector and historical cases, and selecting cases that exceed a threshold (e.g., 0.85) to guide current strategy generation. For example, for the oil pressure gauge scenario, three historical optimal cases with similar salt spray gradients (10-20% / cm) and contact angle change rates (4-6 degrees / hour) may be retrieved.

[0165] In an embodiment of the present application, a coupled feature vector is used as a query condition to perform a nearest neighbor search in a case library of a large language model, and successful compensation cases whose similarity exceeds a preset threshold are returned.

[0166] Step 603: Based on the case set, calculate the predicted value of the droplet distribution density and the predicted value of the salt crystal deposition thickness on the surface of the marine instrument anti-corrosion coating under the current environmental conditions.

[0167] In step 603, the current environmental conditions refer specifically to the actual environmental parameters detected during the current identification task (such as instantaneous salt spray concentration and humidity), which are used to dynamically generate predictions. The predicted values ​​for droplet distribution density and salt crystal deposition thickness refer to the distribution of interferences that may form in the current environment based on historical case studies.

[0168] In the embodiment of the present application, linear deduction is performed based on the environmental material relationship in the matching case and combined with the current parameter differences to calculate the predicted distribution of droplets and salt crystals.

[0169] Step 604: Generate corresponding light source wavelength adjustment suggestion information and polarization angle adjustment suggestion information according to the droplet distribution density prediction value.

[0170] In step 604 , the light source wavelength adjustment suggestion information and the polarization angle adjustment suggestion information refer to optical parameter optimization solutions required for predicting droplet distribution.

[0171] In the embodiment of the present application, based on the droplet density distribution characteristics, the wavelength of the light source that can highlight the contrast of the droplet edge is selected, and the optimal polarization angle for suppressing the reflection of the droplet surface is calculated.

[0172] Step 605: Generate corresponding filter transmittance curve adjustment suggestion information based on the salt crystal deposition thickness prediction value.

[0173] In step 605 , the filter transmittance curve adjustment suggestion information refers to a filtering solution designed based on the salt crystal deposition characteristics.

[0174] In the embodiment of the present application, the effect of salt crystal thickness on scattered light is analyzed to generate a transmittance curve that can effectively filter scattered noise.

[0175] Step 606: Generate an optical compensation strategy by integrating the light source wavelength adjustment suggestion information, the polarization angle adjustment suggestion information, and the filter transmittance curve adjustment suggestion information.

[0176] In the embodiment of the present application, wavelength, polarization and filter suggestions are subjected to conflict detection and parameter fusion to generate a final compensation instruction set.

[0177] Here's a specific example:

[0178] For example, using the engine tachometer's "1500 RPM" detection, the coating contact angle was measured to decrease from an initial 125 degrees at a rate of 8 degrees per hour (the rate of change was calculated from three consecutive measurements of 125, 119, and 112 degrees). Simultaneously, the salt spray concentration gradient increased by 20% per centimeter (calculated from the concentration difference between the upper and lower sensors). These two parameters were bound by timestamp to generate a coupled feature vector. Two similar cases (with a similarity of 0.88) were retrieved from the large language model case database: Case 1 used a 455nm wavelength and a 55° polarization angle, and Case 2 used a 440nm wavelength and a 65° polarization angle. Based on the case data, the current droplet distribution density was predicted to be 15-18 per square millimeter (a weighted average of 12 cases from Case 1 and 20 cases from Case 2), and the salt crystal deposition thickness was 0.05 mm (calculated by interpolation of the concentration gradient). It is recommended to select a wavelength of 450nm (the average of 455nm and 440nm) and a polarization angle of 60 degrees (the middle value between 55 degrees and 65 degrees). The filter transmittance curve is set to 85% transmittance at 450nm (calculated based on the inverse relationship between salt crystal thickness and transmittance).

[0179] In the embodiment of the present application, the intelligent generation of the optical compensation strategy is realized through the above steps, which effectively solves the problem of difficulty in adapting optical parameters in rainy and foggy environments and improves the stability of character recognition under different environmental conditions.

[0180] In order to solve the problem of poor instrument imaging quality in rainy and foggy environments and further improve the feature differentiation of the interference fringe pattern, in some embodiments, step 103: generating the interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy includes:

[0181] Step 701: Based on the light source wavelength adjustment suggestion information in the optical compensation strategy, adjust the wavelength of the incident light beam of the illumination light source so that the incident light beam excites a scattered characteristic light beam representing the droplet morphology on the surface of the anti-corrosion coating.

[0182] In step 701, the scattered characteristic light beam refers to a light beam formed after the incident light is scattered at the droplets on the surface of the anti-corrosion coating, and its intensity distribution reflects the size and distribution density of the droplets.

[0183] In the embodiment of the present application, a multi-spectral light source is configured according to the wavelength adjustment suggestion so that the wavelength of the output light beam is adapted to the current droplet curvature, thereby exciting a scattered light beam with obvious edge features on the coating surface.

[0184] Step 702: Based on the polarization angle adjustment suggestion information in the optical compensation strategy, adjust the polarization angle of the circularly polarized light source to suppress stray reflected light caused by salt crystal deposition on the surface of the anti-corrosion coating of the marine instrument.

[0185] In step 702, stray reflected light refers to interference light formed by disordered reflection on the surface of salt crystal particles, and its polarization characteristics are different from those of the target signal.

[0186] In the embodiment of the present application, the transmission axis direction of the circular polarizer is rotated according to the polarization angle suggestion, so as to selectively suppress the reflected light of the salt crystals in a specific direction and retain the useful refraction signal.

[0187] Step 703: Based on the filter transmittance curve adjustment suggestion information in the optical compensation strategy, adjust the filter transmittance curve of the filter array to enhance the interference contrast between the scattered characteristic light beam and the refracted characteristic light beam.

[0188] In step 703, the refracted characteristic beam is formed by the refraction of the adjusted incident light beam as it passes through the droplets on the surface of the anti-corrosion coating. When the incident light beam illuminates the droplets at a specific wavelength and polarization angle, the light beam is refracted at the droplet-air interface. The refraction angle and phase change carry information about the droplet morphology, forming a refracted characteristic beam with a specific propagation direction. Interference contrast refers to the difference in brightness of the fringes formed by the interference of scattered light and refracted light, which affects the discernibility of features.

[0189] In an embodiment of the present application, the transmittance of the multi-band filter is dynamically configured according to the filter adjustment suggestion, thereby enhancing the target characteristic band while suppressing the noise band.

[0190] Step 704: forming an interference light intensity distribution through the interaction between the scattered characteristic light beam and the refracted characteristic light beam at the imaging plane, and generating an interference fringe pattern according to the interference light intensity distribution.

[0191] In step 704, the interference light intensity distribution refers to the spatial intensity variation pattern formed by the superposition of two characteristic light beams on the imaging plane. The interference fringe pattern generation process involves using an image sensor to capture the spatial intensity distribution of the light formed by the interference of the scattered and refracted characteristic light beams on the imaging plane. The interference light intensity value at each pixel is quantized into grayscale levels, generating a two-dimensional grayscale distribution pattern that simultaneously encodes the scattered intensity and refracted phase differences. In a specific embodiment, when identifying a marine engine tachometer displaying "RPM 1500," the characteristic light beam refracted by a liquid droplet creates concentric interference fringes around the edges of the digits, while the characteristic light beam scattered by salt crystals produces radial fringes around the unit symbol "RPM." In the resulting interference fringe pattern, the vertical strokes of the numeral "1" appear as evenly spaced parallel fringes, while the arc portion of the letter "P" displays gradually curved fringes. This characteristic distribution directly corresponds to the geometric structure of the tachometer characters.

[0192] In an embodiment of the present application, a high-sensitivity image sensor is used to capture light intensity changes in the interference area, and the spatially distributed intensity information is quantified into a grayscale image to generate an interference pattern containing the characteristics of droplets and salt crystals.

[0193] Here's a specific example:

[0194] Taking the identification of "1500 RPM" on an engine tachometer as an example, the optical compensation strategy sets the light source wavelength to 450nm (calculated from the average droplet curvature radius of 4mm using the formula: optimal wavelength = curvature radius × refractive index difference / 2). The polarization angle is adjusted to 60 degrees (the optimal suppression angle determined through experiments on salt crystal reflection properties), and the filter transmittance at 450nm is set to 90% (determined based on a curve showing the relationship between salt crystal thickness (0.1mm) and transmittance). After compensation, the incident light beam generates a clear scattered characteristic beam on the meter surface (parallel fringes with a spacing of 2.1mm are formed at the numeral "1"; fringe spacing = wavelength × droplet curvature radius / refractive index difference). Polarization filtering also effectively suppresses stray light from the salt crystals surrounding the letter "R." The refracted characteristic beam produces a circular interference fringe (radius 3.2mm, calculated from the droplet curvature radius and the incident angle) at the numeral "0," forming a pattern of interference. The refractive index difference is 1.33 (the refractive index difference between water and air), and the 2.3-fold improvement factor is obtained by comparing the signal-to-noise ratio values ​​of the images before and after compensation.

[0195] In the embodiment of the present application, the above steps are used to achieve efficient optical extraction of rain and fog environment characteristics, and the generated interference fringe pattern improves the characteristic distinction between droplets and salt crystals, providing high-quality input for subsequent character recognition.

[0196] Figure 2 A structural diagram of a rain and fog weather instrument character recognition system that integrates large model prior knowledge is provided in an embodiment of the present application, such as Figure 2 As shown, the system includes:

[0197] The acquisition module 21 is used to acquire surface hydrophobicity characteristic data and environmental dynamic parameters of the anti-corrosion coating of the navigation instrument when the navigation instrument is in a rainy and foggy environment.

[0198] The correlation analysis module 22 is used to perform correlation analysis on the surface hydrophobicity characteristic data and the environmental dynamic parameters through a pre-trained large language model to generate an optical compensation strategy.

[0199] The first generating module 23 is configured to generate an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy.

[0200] The second generating module 24 is configured to perform dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively.

[0201] The spatial association module 25 is used to spatially associate the noise suppression map with the distortion correction map to obtain a spatial association map, and iteratively optimize the spatial association map in combination with the character stroke continuity prior rules in the preset large model prior knowledge base to generate instrument character recognition results.

[0202] Figure 2 The rain and fog weather instrument character recognition system that integrates large model prior knowledge can perform Figure 1 The implementation principles and technical effects of the rain and fog weather instrument character recognition method integrated with large-scale model prior knowledge described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the rain and fog weather instrument character recognition system integrated with large-scale model prior knowledge in the aforementioned embodiment has been described in detail in the relevant embodiments of the method and will not be further elaborated here.

[0203] In one possible design, Figure 2 The rain and fog weather instrument character recognition system of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0204] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0205] The processing component 32 is as follows Figure 1 The embodiment provides a method for character recognition of instruments in rainy and foggy weather conditions that integrates large-scale model prior knowledge.

[0206] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0207] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0208] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0209] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0210] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0211] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0212] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a method for character recognition of instruments in rainy and foggy weather conditions that integrates large model prior knowledge.

[0213] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0214] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0215] Through the description of the above embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0216] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for character recognition of rain and fog weather instruments integrating large model prior knowledge, characterized in that: include: Obtain surface hydrophobicity data and environmental dynamic parameters of the anti-corrosion coating of marine instruments in rain and fog environments; Performing correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy; generating an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy; Performing dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively; Spatially correlating the noise suppression map with the distortion correction map to obtain a spatial correlation map, and iteratively optimizing the spatial correlation map based on a priori rules for character stroke continuity in a preset large model prior knowledge base to generate an instrument character recognition result; The correlative analysis of the surface hydrophobicity data and the environmental dynamic parameters by using a pre-trained large language model to generate an optical compensation strategy includes: Binding the coating contact angle change rate in the surface hydrophobicity characteristic data and the salt spray concentration gradient in the environmental dynamic parameters in a time series manner to generate a coupling characteristic vector of the environment and the material; Retrieving matching historical optical compensation cases in the large language model based on the coupling feature vector, and extracting a set of cases whose similarity with the current coupling feature vector reaches a preset similarity threshold from the historical optical compensation cases; Based on the case set, calculate the predicted value of droplet distribution density and salt crystal deposition thickness on the surface of the anti-corrosion coating of the marine instrument under the current environmental conditions; generating corresponding light source wavelength adjustment suggestion information and polarization angle adjustment suggestion information according to the droplet distribution density prediction value; generating corresponding filter transmittance curve adjustment suggestion information according to the salt crystal deposition thickness prediction value; An optical compensation strategy is generated by integrating the light source wavelength adjustment suggestion information, the polarization angle adjustment suggestion information, and the filter transmittance curve adjustment suggestion information.

2. The method according to claim 1, characterized in that The iterative optimization of the spatial association graph in combination with the character stroke continuity prior rules in the preset large model prior knowledge base to generate instrument character recognition results includes: Extracting candidate line segments related to the stroke directions of instrument characters from the spatial association map; Matching the candidate line segments with the character stroke continuity priori rule to generate a candidate line segment set; Iteratively optimizing the spatial correlation map according to the set of candidate line segments to generate an optimized spatial correlation map; Candidate character boundaries that match the predefined character structure template in the character stroke continuity prior rule are extracted from the optimized spatial association map to generate an instrument character recognition result.

3. The method according to claim 2, characterized in that The matching of the candidate line segments with the character stroke continuity priori rule to generate a set of candidate line segments includes: Dividing a set of line segment direction types according to the extension direction and length ratio of the candidate line segments in the spatial association map; Extracting a stroke turning angle range corresponding to the line segment direction type set from the character stroke continuity prior rule; Grouping the candidate line segments into candidate line segment clusters according to corresponding extension directions; Comparing the distance between the start and end points of adjacent line segment pairs in the candidate line segment cluster with a preset distance tolerance, where the start and end point distance is the distance between the start end point of one candidate line segment and the end end point of the other candidate line segment in the adjacent line segment pair; For each pair of adjacent line segments whose distance between the first and last endpoints is less than the distance tolerance, calculating the angle fluctuation amplitude of the adjacent line segment pair based on the stroke turning angle range; Based on the angle fluctuation amplitude, a candidate line segment set is generated.

4. The method according to claim 3, characterized in that The generating of a set of candidate line segments based on the angle fluctuation amplitude includes: Extracting a curvature change allowable threshold corresponding to the line segment direction type set from the character stroke continuity prior rule; Eliminating adjacent line segment pairs whose angle fluctuation amplitude exceeds the curvature change allowable threshold from the candidate line segment cluster to generate a target line segment cluster; Taking the center point of the target line segment cluster as a reference, extending a virtual connecting line along the direction of curvature change, and determining whether the virtual connecting line is consistent with the stroke extension trend of the character structure in the character stroke continuity prior rule; When it is determined to be consistent, dynamically adjusting the angle threshold of adjacent line segments in the target line segment cluster; Adjacent line segment pairs that meet the dynamically adjusted angle threshold are connected and integrated to generate a set of candidate line segments.

5. The method according to claim 2, characterized in that The iteratively optimizing the spatial correlation map according to the set of candidate line segments to generate an optimized spatial correlation map includes: According to the spatial distribution relationship between the endpoints of the line segments in the candidate line segment set, a pixel confidence enhancement region associated with the character stroke overlap probability is divided in the spatial association map; Based on the stroke connection topology constraint relationship in the character stroke continuity prior rule, bidirectional weighted diffusion is performed on the pixel confidence enhancement area to construct a confidence distribution network covering the complete character stroke outline; According to the confidence distribution network and the set of candidate line segments, the confidence levels of the pixels in the spatial association map are dynamically enhanced or attenuated to generate an optimized spatial association map.

6. The method according to claim 1, characterized in that Generating an interference fringe pattern corresponding to the surface of the anti-corrosion coating of a marine instrument according to the optical compensation strategy includes: Based on the light source wavelength adjustment suggestion information in the optical compensation strategy, the wavelength of the incident light beam of the illumination light source is adjusted so that the incident light beam excites a scattered characteristic light beam representing the morphology of the droplets on the surface of the anti-corrosion coating; Based on the polarization angle adjustment suggestion information in the optical compensation strategy, the polarization angle of the circularly polarized light source is adjusted to suppress stray reflected light caused by salt crystal deposition on the surface of the anti-corrosion coating of the marine instrument; Adjusting the filter transmittance curve of the filter array based on the filter transmittance curve adjustment suggestion information in the optical compensation strategy to enhance the interference contrast between the scattered characteristic light beam and the refracted characteristic light beam; An interference light intensity distribution is formed by the interaction between the scattered characteristic light beam and the refracted characteristic light beam in the imaging plane, and an interference fringe pattern is generated according to the interference light intensity distribution.

7. A rain and fog weather instrument character recognition system integrating large model prior knowledge, characterized by: include: An acquisition module is used to obtain surface hydrophobicity characteristic data and environmental dynamic parameters of the anti-corrosion coating of the navigation instrument when the navigation instrument is in a rainy and foggy environment; A correlation analysis module, configured to perform correlation analysis on the surface hydrophobicity data and the environmental dynamic parameters using a pre-trained large language model to generate an optical compensation strategy; A first generating module is used to generate an interference fringe pattern corresponding to the surface of the anti-corrosion coating of the marine instrument according to the optical compensation strategy; A second generating module is used to perform dual-branch processing on the interference fringe pattern to generate a noise suppression pattern and a distortion correction pattern respectively; a spatial association module for spatially associating the noise suppression map with the distortion correction map to obtain a spatial association map, and iteratively optimizing the spatial association map based on a priori rules for character stroke continuity in a preset large model prior knowledge base to generate an instrument character recognition result; The correlative analysis of the surface hydrophobicity data and the environmental dynamic parameters by using a pre-trained large language model to generate an optical compensation strategy includes: Binding the coating contact angle change rate in the surface hydrophobicity characteristic data and the salt spray concentration gradient in the environmental dynamic parameters in a time series manner to generate a coupling characteristic vector of the environment and the material; Retrieving matching historical optical compensation cases in the large language model based on the coupling feature vector, and extracting a set of cases whose similarity with the current coupling feature vector reaches a preset similarity threshold from the historical optical compensation cases; Based on the case set, calculate the predicted value of droplet distribution density and salt crystal deposition thickness on the surface of the anti-corrosion coating of the marine instrument under the current environmental conditions; generating corresponding light source wavelength adjustment suggestion information and polarization angle adjustment suggestion information according to the droplet distribution density prediction value; generating corresponding filter transmittance curve adjustment suggestion information according to the salt crystal deposition thickness prediction value; An optical compensation strategy is generated by integrating the light source wavelength adjustment suggestion information, the polarization angle adjustment suggestion information, and the filter transmittance curve adjustment suggestion information.

8. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a rain and fog weather instrument character recognition method that integrates large model prior knowledge as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the method for character recognition of rain and fog weather instruments integrating large model prior knowledge as described in any one of claims 1 to 6 is implemented.

Images