Organic coating corrosion visual enhancement and robust recognition method for complex environment

Abstract

The present application belongs to the technical field of intelligent detection of material corrosion, and relates to a visual enhancement and robust identification method for organic coating corrosion in complex environments. The method adopts an environment-driven enhancement strategy to dynamically adjust image enhancement processing parameters according to the synchronously collected environmental parameters (illumination intensity, salt spray deposition degree, humidity, etc.) for self-adaptive illumination normalization, contrast enhancement and noise suppression of the coating image in view of the interference factors such as illumination variation, salt spray attachment and surface pollution in the actual engineering environment. On this basis, the environmental parameter information is introduced to correct the visual features, and the robust identification of the corrosion state of the organic coating is realized in combination with an electrochemical constraint identification model. The method can effectively reduce the influence of environmental interference on the identification result, improve the stability and accuracy of corrosion identification, and is suitable for organic coating corrosion detection in complex service environments.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection and engineering application technology of material corrosion, and in particular to a visual enhancement and robust identification method for organic coating corrosion in complex environments. Background Technology

[0002] With the increasingly widespread application of image-based intelligent recognition technology in corrosion detection, its performance in real-world engineering environments is receiving increasing attention. Unlike the ideal conditions of a laboratory, acquiring images of coating surfaces in real-world service environments faces numerous challenges. First, lighting conditions are complex and variable. Outdoor structures experience significant differences in the intensity, angle, and color temperature of natural light at different times and under different weather conditions. Even with auxiliary light sources, it is difficult to completely eliminate the effects of shadows, reflections, and uneven lighting. Second, salt spray deposition is prevalent in marine atmospheric environments. The white crystals or liquid films formed by salt spray on coating surfaces can cover or alter the original corrosion morphology, appearing as blurred white spots or reflective areas in images, easily confused with corrosion products. Furthermore, dust pollution in industrial areas or windy, sandy environments, as well as accidental adhesions such as bird droppings and oil stains, can also severely interfere with image quality.

[0003] Existing image-based corrosion visual recognition methods rely heavily on training data collected in controlled laboratory environments or carefully selected field samples, resulting in relatively ideal image quality. However, when these methods are directly deployed to complex field environments, the aforementioned environmental interference factors can lead to degraded image quality and distortion of visual features, causing a sharp decline in the performance of the recognition model. For example, a model that performs well on laboratory data may misidentify a normal surface as corroded under strong field light or miss genuine corrosion points in areas with severe salt spray. This sensitivity to environmental interference severely limits the reliable application of intelligent recognition technology in engineering practice.

[0004] Therefore, it is necessary to develop a visual enhancement and robust recognition method that can actively perceive and adapt to complex environmental interference. Through image preprocessing and feature correction, the negative impact of environmental factors on the recognition results can be eliminated or reduced, thereby ensuring the stability and accuracy of the corrosion recognition system under real working conditions. Summary of the Invention

[0005] The purpose of this invention is to provide a visual enhancement and robust identification method for organic coating corrosion in complex environments. It aims to overcome the limitations of existing visual recognition methods, such as sensitivity to environmental interference and poor on-site adaptability. Through the technical path of "environmental perception + environment-driven enhancement + feature correction", it achieves highly robust identification of the corrosion state of organic coatings.

[0006] To achieve the above objectives, the core idea of ​​this invention is to simultaneously record key environmental parameters during image acquisition, using this as prior information to dynamically drive adaptive image enhancement processing (i.e., adopting different processing strategies and parameters according to different values ​​of environmental parameters), and further modifying the extracted visual features to reflect the true corrosion state rather than environmental interference. Specifically, this invention is achieved through the following technical solution: The first step is to acquire image data of the organic coating surface and corresponding environmental parameters. At the testing site, a smart data acquisition terminal equipped with environmental sensors is used to photograph the target measurement point. At the moment the image is captured, the terminal simultaneously records the light intensity, ambient humidity, and parameters characterizing the degree of salt spray deposition at that location (such as air salinity or a salt spray coverage index estimated from a local area of ​​the image). These environmental parameters will correspond one-to-one with the image data, serving as crucial information for subsequent processing.

[0007] The second step involves environment-driven adaptive enhancement of the image data. This step aims to eliminate or mitigate environmental interference at the signal level. Its core lies in dynamically adjusting key parameters in the processing algorithm based on different environmental parameters to achieve environment-driven adaptive enhancement.

[0008] Illumination Normalization: Based on the synchronously recorded illumination intensity values, the target brightness parameter for brightness normalization is dynamically adjusted. For example, when the illumination intensity is higher than 1000 Lux, the target brightness is set to a lower value (e.g., 120) to avoid overexposure; when the illumination intensity is lower than 200 Lux, the target brightness is set to a higher value (e.g., 180) to compensate for underexposure, thereby uniformly mapping the image to a standard illumination space and eliminating brightness differences caused by different shooting times and weather conditions.

[0009] Contrast Enhancement: The threshold parameter for contrast enhancement is dynamically adjusted based on the ambient humidity level. When the humidity is high (>85%), the image tends to appear hazy. In this case, a lower limit contrast threshold (e.g., 1.5) is used for adaptive histogram equalization to avoid excessive noise enhancement. When the humidity is moderate, a standard threshold (2.0) is used.

[0010] Noise Suppression: For specific "structural noise" such as salt spray adhesion and surface contamination, the filtering intensity parameters of the denoising algorithm are dynamically adjusted based on the salt spray deposition index. When the salt spray deposition index is high (e.g., the proportion of bright areas > 20%), a stronger dehazing process is used (e.g., setting the window radius in the dark channel prior algorithm to a larger value); when the index is low, conventional denoising or only slight processing is used to preserve details. For dust contamination, the kernel size of morphological operations is dynamically adjusted based on the estimated contamination area to achieve targeted repair.

[0011] The adaptive enhancement driven by the above environmental parameters makes the image processing process more closely match the actual on-site environment and improves the enhancement effect.

[0012] The third step is to extract the visual feature parameters of the enhanced image. For the enhanced image, quantitative visual features that can characterize the corrosion state of the coating are extracted. These features may include color distribution features (such as the mean and variance of each color channel), texture statistical features (such as energy, entropy, and contrast based on the gray-level co-occurrence matrix), and surface defect distribution features (such as the proportion of rust spots and blister density obtained through semantic segmentation).

[0013] The fourth step involves correcting the visual features based on environmental parameters to reduce the impact of environmental interference. This is the core innovation of this invention. After image enhancement, some environmental interference is suppressed, but some complex and nonlinear interference may still remain in the features. Therefore, this step introduces environmental parameters to perform secondary correction on the initially extracted visual features. The theoretical basis is that there is a learnable mapping relationship between environmental parameters (such as light intensity and humidity) and visual features (such as color mean and texture contrast). Specifically, a lightweight correction model (such as a three-layer fully connected network) is established, taking the environmental parameters as input and outputting a correction vector for the original visual features. The original features are then added to the correction vector to obtain the final environment-invariant features. This correction model is trained by minimizing the feature differences of the same measurement point under interference and standard environments, thus achieving "different interpretations of features in different environments." The corrected visual features are closer to the ideal features of the measurement point under standard environments and can more realistically reflect the corrosion state of the coating itself.

[0014] The fifth step involves inputting the corrected visual features into the corrosion recognition model for identification, outputting the corrosion state result. The environmentally corrected visual features possess better environmental invariance and are input into a trained, high-precision corrosion recognition model. This model preferably adopts the one constructed in the inventor's prior patent (Patent 3: A Visual Recognition Method for Organic Coating Corrosion Based on Electrochemical Constraints), as it inherently possesses physical constraints and exhibits higher robustness to feature quality. After forward computation, the model ultimately outputs the corrosion level (e.g., 0-3 or 0-4) and corresponding corrosion state description information for the measurement point.

[0015] The beneficial effects of this invention are as follows: First, by introducing environmental parameter sensing, it achieves proactive identification and targeted processing of environmental interference, overcoming the limitations of traditional "one-size-fits-all" image enhancement methods. Second, it constructs a dual anti-interference mechanism of "environment-driven enhancement + feature correction," where environment-driven enhancement dynamically adjusts processing parameters according to the actual environment, eliminating interference more accurately at the signal layer; feature correction further compensates for residual interference at the feature layer, significantly improving the robustness of the identification method. Third, this method is naturally compatible with electrochemically constrained identification models, fully leveraging the advantages of physical constraint models in feature robustness to form a complete, practically-oriented intelligent corrosion identification solution. Fourth, this method is particularly suitable for complex outdoor environments such as marine atmospheres, industrial pollution, and drastic changes in lighting, significantly improving the stability and accuracy of intelligent detection systems under harsh conditions. Finally, this method does not change the operating habits of frontline personnel; it only requires adding environmental sensors and corresponding software algorithms to the acquisition terminal, achieving seamless migration from the laboratory to the field, possessing extremely high engineering practical value and promising prospects for widespread application. Attached Figure Description

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the overall process of the present invention for visual enhancement and robust recognition of organic coating corrosion in complex environments, showing the entire process from synchronous acquisition of environmental parameters and images, to image enhancement, feature extraction, feature correction, and finally output of corrosion results by the recognition model. Figure 2 This diagram illustrates the key steps of image enhancement processing in this invention, exemplarily showing a comparison of the effects of the original image, the image after illumination normalization, the image after contrast enhancement, and the image after salt spray noise suppression. Figure 3 This is a schematic diagram of the visual feature correction mechanism based on environmental parameters in this invention, illustrating how environmental parameters act on the original visual features through a correction model to generate environmentally invariant features. Detailed Implementation

[0017] The present invention will be further explained below with reference to specific implementation schemes, but it is not limited to the present invention. The structures, proportions, sizes, etc. shown in the accompanying drawings are only used to complement the content disclosed in the specification, so as to enable those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modification of the structure, change of the proportion relationship or adjustment of the size, without affecting the effect and purpose that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to two specific embodiments. Embodiment 1 simulates typical environmental interference in a laboratory environment, verifying the effectiveness of each module of the method of this invention through the controlled variable method, particularly the role of the parameter adjustment mechanism in the environment-driven enhancement strategy. Embodiment 2 conducts long-term tracking tests on a steel structure communication tower in the actual environment of the coastal area of ​​Hainan Province, my country, comprehensively evaluating the robustness and practicality of the method of this invention in real extreme environments.

[0019] Example 1: Simulation and Algorithm Verification of Laboratory Environmental Interference Simulation conditions settings.

[0020] This embodiment aims to simulate three common environmental disturbances in a controlled laboratory environment: light variation, salt spray adhesion, and surface dust contamination, in order to quantitatively evaluate the performance of the method of this invention. The same 2A12 aluminum alloy / epoxy polyurethane coated samples as those in Patent 3 were selected. These samples have undergone accelerated corrosion testing covering different corrosion states from level 0 to 3. The samples were fixed on a six-axis platform with adjustable illumination, placed inside a high and low temperature humidity chamber, with a precision salt spray nozzle positioned in front of the samples. Four imaging conditions were designed for the experiment: standard conditions (light intensity 500 Lux, no interference), strong light conditions (light intensity 2000 Lux, simulating direct midday sunlight), salt spray adhesion conditions (under standard illumination, the sample surface is sprayed with 5wt% NaCl solution for 30 seconds and then left to stand for 2 minutes, simulating salt spray deposition), and dust contamination conditions (under standard illumination, the sample surface is uniformly sprayed with standard test dust, simulating windblown sand or industrial dustfall). Under each condition, images were acquired for all samples with different corrosion levels, resulting in a total of 160 sets of image data: 4 (conditions) × 4 (levels) × 10 (parallel samples). Simultaneously, during each acquisition, corresponding environmental parameters—light intensity, ambient humidity, and salt spray deposition index—were recorded using a illuminometer, hygrometer, and image analysis (calculating the proportion of bright areas in the salt spray image).

[0021] Environment-driven enhancement and feature extraction.

[0022] For the 160 sets of images acquired, the environment-driven enhancement processing flow of this invention was applied, and the processing parameters were dynamically adjusted according to the environmental parameters: Illumination normalization: Based on the recorded illumination intensity L, set the target brightness T: if L > 1000 Lux, T = 120; if L < 200 Lux, T = 180; otherwise, T = 150. Adjust the overall image brightness to T.

[0023] Contrast enhancement: Set the contrast limit threshold C of CLAHE according to the humidity H: if H>85%, C=1.5; otherwise C=2.0.

[0024] Noise suppression: For salt spray images, adjust the window radius r of the dark channel dehazing algorithm according to the salt spray deposition index D (proportion of bright areas): if D > 20%, r = 15; otherwise, r = 9. For dust pollution images, adjust the kernel size k of the morphological opening operation according to the pollution area proportion P: if P > 10%, k = 5; otherwise, k = 3.

[0025] After enhancement, 17-dimensional handcrafted visual features were extracted from each image, including color features, texture features, and defect distribution features. Simultaneously, for comparison, the same features were also extracted from the original, unenhanced images.

[0026] Feature correction and model recognition.

[0027] A baseline feature space was established based on features extracted from images under all standard conditions. For images under conditions of strong light, salt spray, and dust, the original 17-dimensional features extracted after enhancement, along with synchronously recorded environmental parameters (light intensity, humidity, and salt spray deposition index), were input into a pre-trained lightweight correction network. This network is a three-layer fully connected network with a 20-dimensional input (17-dimensional features + 3-dimensional environmental parameters) and an output of a 17-dimensional correction vector. The original features and the correction vector were added to obtain the final environmentally corrected features. The training objective of the correction network was to minimize the mean square error between the corrected features and the baseline features of the same sample under standard conditions. The corrected features were then input into the electrochemically constrained corrosion recognition model trained in Patent 3, which outputs the corrosion level. For comparison, the original image features without enhancement and the image features that were only enhanced but not corrected were also input into the same recognition model.

[0028] Results and Analysis.

[0029] The recognition results are shown in the table below: Under strong light conditions, the recognition accuracy of the unprocessed image is only 70%. After environmental-driven enhancement according to this invention, the accuracy increases to 82% (an improvement compared to the 80% of ordinary enhancement, proving the effectiveness of the parameter adjustment mechanism). After feature correction, the final accuracy reaches 85%. Under salt spray conditions, the unprocessed accuracy is 65%, which increases to 80% after environmental-driven enhancement (compared to 78% with ordinary enhancement) and reaches 83% after feature correction. Under dust pollution conditions, the unprocessed accuracy is 68%, which increases to 82% after environmental-driven enhancement (compared to 80% with ordinary enhancement) and reaches 84% ​​after feature correction. Statistical results show that the dual mechanism of "environment-driven enhancement + feature correction" proposed in this invention can significantly improve the recognition accuracy of the model under various environmental interferences. Compared with the unprocessed image, the average accuracy is improved by about 15-18 percentage points. In particular, the environmental-driven enhancement module, due to dynamically adjusting the processing parameters according to environmental parameters, brings an additional gain of 2-3 percentage points compared to ordinary enhancement, proving the effectiveness of the parameter adjustment mechanism. The feature correction module provides an additional 3-5 percentage points of gain on top of image enhancement, demonstrating its effectiveness in eliminating residual interference in the feature layer. This embodiment rigorously verifies the functionality of each module and the overall effect of the method of this invention through laboratory simulation.

[0030] Example 2: Application of a steel structure tower project in Hainan Application environment and challenges.

[0031] This embodiment selects a 50-meter-high steel structure integrated communication tower on the coast of Hainan Province, my country, as the engineering verification object. The coastal area of ​​Hainan Province has a typical hot and humid maritime climate, with an average annual temperature of 28℃, relative humidity consistently above 85%, and salt spray deposition as high as 2.5 mg / (cm³). 2 •h). The tower surface coating system consists of epoxy zinc-rich primer + epoxy micaceous iron oxide intermediate paint + acrylic polyurethane topcoat, and has been in service for 5 years. On-site identification faces three major extreme challenges: First, the lighting is extremely variable, ranging drastically from strong direct sunlight at noon (>100,000 Lux) to low light on cloudy or rainy days (<50 Lux); second, severe salt spray deposition, with the tower surface often covered with a layer of white salt frost visible to the naked eye; and third, after typhoons, the surface is often covered with plant debris and mud blown by the wind.

[0032] Data acquisition and processing.

[0033] The project team selected 20 fixed monitoring points at different heights (10 meters, 25 meters, and 40 meters) and orientations (windward and leeward) on the tower, and conducted a six-month tracking test (spanning the typhoon and dry seasons). Each test involved two personnel traveling to the monitoring point in an aerial work platform. A handheld intelligent data acquisition terminal integrating a illuminometer, temperature and humidity sensors, and an industrial camera was used to continuously capture five images at each monitoring point, simultaneously recording environmental parameters (light intensity, humidity, and salt spray coverage index estimated locally from the images). To obtain a reference standard for the actual corrosion state, after each image acquisition, a portable solid gel electrolyte probe was used to measure electrochemical impedance spectroscopy at the same location to obtain the low-frequency impedance modulus. A senior corrosion engineer accompanying the team then comprehensively assessed the corrosion level based on the images and measured data. Approximately 400 sets of valid data samples were collected throughout the cycle, covering all stages from Grade 0 (intact) to Grade 3 (severe corrosion), and various environmental conditions including strong light, weak light, cloudy days, after rain, and salt spray adhesion.

[0034] Identification process and effectiveness evaluation.

[0035] Four hundred sets of image data collected on-site were batch-processed according to the processing flow of this invention. First, using the light intensity, humidity, and salt spray coverage index from the image metadata, and employing the parameter adjustment rules summarized in Example 1, environment-driven adaptive enhancement was performed on each image. Then, 17-dimensional visual features of the images were extracted and input into the correction network along with synchronously recorded environmental parameters to obtain environment-corrected features. Finally, the corrected features were input into the recognition model of Patent 3 to output the corrosion level. The model's output was compared with the corrosion level comprehensively assessed by on-site experts. As a comparison, a commercial visual recognition software trained only on standard laboratory data without any enhancements or corrections was deployed on another terminal to recognize the same batch of images.

[0036] Results and Engineering Effects Analysis.

[0037] Statistical results show that the method of this invention achieved an average recognition accuracy of 86.2% across 400 test samples. Specifically, the accuracy was 84.5% for samples under extreme lighting conditions (strong light or extreme darkness), 83.1% for samples with significant salt spray adhesion, and remained above 80% even on complex samples with multiple interferences. In contrast, the average accuracy of comparable commercial software was only 72.3%, and its accuracy plummeted to around 60% under strong light and salt spray conditions, exhibiting significant performance fluctuations. Field inspection personnel reported that the method of this invention operates smoothly, with a single recognition time of approximately 15 seconds (including image acquisition and real-time processing), and provides stable and reliable recognition results under various weather conditions. Particularly noteworthy was its effectiveness in post-typhoon inspections, where tower surfaces were covered with a large amount of contaminants, making traditional visual inspection difficult. This method, through image enhancement and feature correction, successfully identified several locations covered in contaminants but already showing early corrosion, providing crucial information for timely repairs. The data collected over six months also revealed a high degree of consistency between the model identification results and the measured electrochemical impedance values, further validating the reliability of the method.

[0038] Summary of project results.

[0039] The successful application of this embodiment in the extremely complex environment of the Hainan coast fully demonstrates the excellent environmental adaptability and robustness of the method of the present invention. Through environmental parameter-driven adaptive enhancement combined with a feature correction mechanism, this method can maintain high-precision corrosion identification capabilities under actual working conditions such as drastic changes in light intensity, severe salt spray, and surface contamination. It provides a practical and feasible technical means for intelligent operation and maintenance of critical infrastructure in harsh environments such as islands and coastlines, and has significant military and civilian value.

[0040] In summary, the present invention provides a visual enhancement and robust identification method for organic coating corrosion in complex environments. By organically combining environmental perception and feature correction, it effectively solves the identification stability problem caused by environmental interference in practical engineering applications, and significantly improves the practical application level of intelligent corrosion detection technology.

[0041] Matters not covered in this invention are common knowledge.

[0042] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for visual enhancement and robust recognition of organic coating corrosion in complex environments, characterized in that, Includes the following steps: (1) Obtain image data of the organic coating surface and corresponding environmental parameter data; (2) Based on different values ​​of environmental parameters, an environment-driven enhancement strategy is adopted to adaptively enhance the image data, including illumination normalization, contrast enhancement and noise suppression. (3) Extract the visual feature parameters of the enhanced image; (4) Correct visual features based on environmental parameters to reduce the impact of environmental interference; (5) Input the corrected visual features into the corrosion recognition model for recognition and output the corrosion status result.

2. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The environmental parameters mentioned include light intensity, humidity, and salt spray deposition level.

3. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The image enhancement includes brightness normalization and contrast stretching.

4. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The noise suppression includes removing salt spray adhesion and pollution interference.

5. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The environment-driven enhancement strategy in step (2) specifically involves: dynamically adjusting the target brightness parameter of the illumination normalization based on the illumination intensity value; dynamically adjusting the filtering intensity parameter of the denoising algorithm based on the salt spray deposition index; and dynamically adjusting the threshold parameter of the contrast enhancement based on the humidity value, thereby achieving adaptive enhancement driven by environmental parameters.

6. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The visual feature correction is achieved through the mapping relationship between environmental parameters and visual features.

7. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1 or 6, characterized in that, The step (4) of correcting visual features based on environmental parameters is achieved by establishing a correction model that takes environmental parameters as input and feature correction amount as output. This model adjusts the original features according to the mapping relationship between environmental parameters and visual features to generate environmentally invariant features.

8. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The identification model is an electrochemical constraint-based identification model.

9. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, The output includes the corrosion level and a description of the corrosion state.

10. The method for visual enhancement and robust recognition of organic coating corrosion in complex environments according to claim 1, characterized in that, It is suitable for identifying corrosion of organic coatings under various environmental conditions, including complex outdoor environments such as marine atmosphere, industrial pollution, and drastic changes in light.

Images