Composite optical detection method and system for metal surface corrosion

Abstract

The invention discloses a composite optical detection method and system for metal surface corrosion, and the method comprises the steps: employing a line scanning camera and a spectrum confocal sensor to carry out the synchronous data collection through a time-space registration method, obtaining the RGB image data of a to-be-detected metal surface through the line scanning camera, i.e., color data, and obtaining the color data of the to-be-detected metal surface; acquiring point cloud data, namely morphology data, of the metal surface to be detected by using a spectral confocal sensor; carrying out multi-modal data fusion on the RGB data and the point cloud data; a data fusion result is optimized by introducing a confidence coefficient weight; a double-threshold segmentation method of a color threshold and a depth threshold is adopted, and a corrosion area in an optimized data fusion result is preliminarily identified; performing morphological post-processing on the preliminary recognition result through opening operation and closing operation to generate a final recognition result; and carrying out quantitative evaluation on the corrosion area by utilizing a final identification result. The method provided by the invention can realize a rapid and accurate detection process.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental effect testing, and in particular relates to a composite optical detection method and system for surface corrosion of metal components. Background Art

[0002] While metal materials are widely used, the corrosion problem is becoming increasingly serious, especially in industries such as aviation, shipbuilding, and energy. The corrosion of metal components will directly affect their performance and safety. When conducting surface corrosion damage detection on metal components, the usual practice is to use a camera to take pictures and then evaluate the damage through the image. The problem with this approach is that for three-dimensional components, only the plane projection image information in a certain direction is retained, and due to the existence of slopes and curved surfaces, the corrosion characteristics will be distorted, and the plane image information cannot fully reflect the true corrosion morphology, area, depth and other information. Researchers conduct corrosion evaluation based on incomplete and inaccurate image data of this information, which results in large systematic errors. If a single optical sensing device (such as laser scanning) is used for detection, only the dimensional information of the metal component surface can be obtained, and the color information is lost. Due to the influence of the detection range and angle, only the local morphology is obtained, lacking global correlation, and it is impossible to distinguish between corrosion types such as pitting / uniform corrosion, and it is impossible to fully obtain multi-dimensional information such as the depth and morphology of the corrosion pit.

[0003] Existing equipment often independently collects topographic and optical information, resulting in data asynchrony in time and space, leading to fusion errors and making it difficult to accurately correlate the geometric characteristics and chemical state of the corroded area. Therefore, a new metal surface corrosion detection method is urgently needed that can efficiently and comprehensively obtain corrosion damage information to meet the needs of accurately detecting and characterizing the corrosion state of metal components. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a composite optical detection method and system for surface corrosion of metal components, which adopts a high-definition line scan camera lens and a spectral confocal displacement sensor to synchronously collect the color, size and depth information of the metal surface, combines with a position encoder to realize spatial alignment of different sensor data, and uses data fusion technology to comprehensively characterize the state and degree of metal surface corrosion, thereby realizing high-precision, efficient and automated metal surface corrosion detection.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] One aspect of the present invention provides a composite optical detection method for metal surface corrosion, comprising:

[0007] Through the spatiotemporal registration method, a line scan camera and a spectral confocal sensor are used for synchronous data acquisition. The line scan camera is used to obtain RGB image data (i.e., color data) of the metal surface to be inspected, and the spectral confocal sensor is used to obtain point cloud data (i.e., topographic data) of the metal surface to be inspected.

[0008] Perform multimodal data fusion on RGB data and point cloud data;

[0009] By introducing confidence weights, the data fusion results are optimized;

[0010] A dual-threshold segmentation method of color threshold and depth threshold is used to preliminarily identify the corrosion area in the optimized data fusion results;

[0011] Through opening and closing operations, the preliminary recognition results are subjected to morphological post-processing to generate the final recognition results;

[0012] The final identification results are used to quantitatively evaluate the corrosion area.

[0013] Furthermore, the spatiotemporal registration method includes:

[0014] Determine the axis distance and scanning speed of the line scan camera and spectral confocal sensor;

[0015] Determine the time delay based on the wheelbase and scanning speed;

[0016] The trigger timing of the line scan camera and the spectral confocal sensor is adjusted according to the time delay, thereby ensuring that the timestamps of the data collected by the line scan camera and the spectral confocal sensor are aligned.

[0017] Furthermore, by introducing confidence weights, the optimized data fusion results include:

[0018] According to the weight adjustment strategy, the color weight of the RGB image data and the morphology weight of the point cloud data are adaptively adjusted. The weight adjustment strategy includes: determining the corresponding weight according to the noise level of the line scan camera and the spectral confocal sensor, adjusting the weight according to the corrosion depth, adjusting the weight according to the image contrast, etc.

[0019] Furthermore, the method also includes enhancing and fusing the edge area of the corrosion area based on corrosion feature extraction.

[0020] Furthermore, preliminary identification of the corrosion area in the optimized data fusion results includes:

[0021] Convert the RGB image data in the optimized data fusion result into HSV image data, and extract the H channel, S channel and V channel information;

[0022] Set the color threshold of H channel, S channel and V channel;

[0023] According to the point cloud data in the optimized data fusion result, the depth information is normalized to generate the reference plane Z base ;

[0024] Using reference plane Z base , generating the corrosion depth threshold;

[0025] Based on the color threshold and corrosion depth threshold, the preliminary identification results of the corrosion area are generated.

[0026] Furthermore, the final identification results are used to quantitatively evaluate the corrosion area, including:

[0027] Based on the final identification results, determine the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature;

[0028] The corrosion index is determined by taking a weighted sum of the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature;

[0029] Use the corrosion index to determine the corrosion level.

[0030] Furthermore, the determination of the curvature of the corrosion morphology includes:

[0031] Obtain the point cloud data of the corrosion area segmented after data fusion;

[0032] Calculate the second derivative of the depth data for each point in the corrosion area;

[0033] The absolute values of the second derivatives of the depth data at each point in the corrosion area are averaged;

[0034] The average value is taken as the corrosion morphology curvature of the corrosion area.

[0035] Furthermore, using the final identification results, a quantitative assessment of the corrosion area also includes a combination of one or more of the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature to determine the corrosion type.

[0036] Another aspect of the present invention provides a composite optical detection system for metal surface corrosion, the system comprising: a three-axis motion control platform, a line scan camera, a spectral confocal sensor and a data processing device.

[0037] The three-axis motion control platform includes three mutually perpendicular motion axes;

[0038] The line scan camera and the spectral confocal sensor are connected to the same motion axis of the three-axis motion control platform, and the wheelbase between the line scan camera and the spectral confocal sensor is fixed. The line scan camera and the spectral confocal sensor are controlled to move along the three motion axes by a servo motor.

[0039] The line scan camera is used to obtain RGB image data, i.e., color data, of the metal surface to be inspected;

[0040] The spectral confocal sensor is used to obtain point cloud data, i.e., topographic data, of the metal surface to be inspected;

[0041] The data processing device is used to execute the composite optical detection method for metal surface corrosion as described above.

[0042] Furthermore, the system further comprises an electric rotating stage, and the integrated sensor of the line scan camera and the spectral confocal sensor can be rotated by a servo motor.

[0043] The beneficial effects of the present invention are:

[0044] By combining a high-definition line scan camera with a spectral confocal displacement sensor, the present invention can simultaneously acquire multi-dimensional information on corrosion, including color, depth, and size. This is more comprehensive and accurate than traditional methods. Furthermore, synchronous acquisition and data fusion technology greatly improve the automation of corrosion detection, reduce manual intervention, and achieve an efficient and rapid detection process. The present invention can characterize multiple parameters of metal surface corrosion, including color, depth, morphology, and size, helping researchers conduct a comprehensive assessment of corrosion damage.

[0045] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0047] Figure 1 It is a composite optical detection system diagram of metal surface corrosion;

[0048] Figure 2 Is the optical test sensor layout

[0049] Figure 3 It is a line scan corrosion detail map;

[0050] Figure 4This is a graph of data collected by the spectral confocal displacement sensor;

[0051] Figure 5 It is a three-dimensional data graph;

[0052] Figure 6 It is a data fusion flow chart;

[0053] Figure 7 This is a data fusion example diagram;

[0054] Figure 8 It is a steel plate corrosion diagram. DETAILED DESCRIPTION

[0055] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the preferred embodiments are only for illustrating the present invention, and are not intended to limit the scope of protection of the present invention.

[0056] The present invention provides a composite optical detection system for metal surface corrosion, which includes: a three-axis motion control platform, a line scan camera, a spectral confocal sensor, and a data processing device. The three-axis motion control platform includes three mutually perpendicular motion axes; the line scan camera and the spectral confocal sensor are connected to the same motion axis of the three-axis motion control platform, and the wheelbase between the line scan camera and the spectral confocal sensor is fixed. A servo motor is used to control the line scan camera and the spectral confocal sensor to move along the three motion axes; the line scan camera is used to obtain RGB image data, i.e., color data, of the metal surface to be detected; the spectral confocal sensor is used to obtain point cloud data, i.e., morphology data, of the metal surface to be detected; and the data processing device is used to execute the composite optical detection method for metal surface corrosion.

[0057] In some embodiments, the system further comprises a motorized rotating stage, and the integrated sensor of the line scan camera and the spectral confocal sensor can be rotated by a servo motor control.

[0058] Figure 1 This is a diagram of a composite optical detection system for metal surface corrosion. Figure 1 In the figure, 1 is a three-axis motion control platform; 2 is a high-definition line scan camera; 3 is a spectral confocal displacement sensor (or simply "spectral confocal sensor"); 4 is a host computer (including data processing equipment); 5 is an electric rotary table; 6 is an electrical cabinet; 7 is a stage; and 8 is a light source.

[0059] The high-definition line scan camera 2 is used to capture images of the metal surface and obtain surface color information and texture features. The spectral confocal displacement sensor 3 is used to scan the metal surface and obtain surface three-dimensional morphology data. The stage 7 is used to carry the sample to be tested, and the three-axis motion control platform 1 is used to realize the accurate positioning and scanning test movement of the high-definition line scan camera 2 and the spectral confocal displacement sensor 3. If the test sample is a plate-shaped knot, it can be placed directly on the stage 7. When the object to be tested is a cylinder, in order to perform a full-scale detection of the side of the cylinder, the cylindrical test object can be placed on an electric rotating stage (i.e., an electric rotating stage). By rotating the electric rotating stage, the object to be tested can be fully detected. At the same time, the spectral confocal sensor can be rotated by the servo motor control, for example, rotated 90° to adapt to the detection of the cylinder. The host computer (data processing equipment) 4 is used to fuse the collected image information and three-dimensional morphology data, extract corrosion characteristic parameters, and evaluate the degree of corrosion.

[0060] Figure 2 shows the layout of the optical test sensors, namely the line scan camera 2 and the spectral confocal sensor 3 .

[0061] The three-axis motion control platform in the above-mentioned inspection system is a desktop gantry-style platform with a horizontal (X-axis) displacement range of 400mm, a front-back (Y-axis) displacement range of 400mm, and a vertical (Z-axis) displacement range of 200mm. Separate servo motors are used to control the X-axis, Y-axis, and Z-axis motions.

[0062] The displacement distance and speed of the line scan camera and / or spectral confocal sensor can be digitally adjusted, for example, the displacement accuracy is better than 0.05 mm, the rotation range is 0° to 90°, and the rotation accuracy is 0.1°.

[0063] As an example, the line scan camera is a 4K color network interface linear array camera with a 35mm large-area lens, a CMOS sensor, a pixel size of 7μm, and a resolution of 4096*2.

[0064] The spectral confocal displacement sensor is a point sensor with a measuring range of ±3mm, a measuring angle of ±14°, a spot diameter of Φ21μm, a lateral resolution of 10.5μm, a linear error of ≤1.2μm, and a minimum measurable thickness of 5mm.

[0065] The host computer is a desktop computer with Win10 operating system and equipped with a high-resolution display.

[0066] The electric rotary table is adjustable from 0 to 50° / s with an adjustment accuracy of 0.1° / s.

[0067] The electrical cabinet is the installation area for PLC, switches, wiring harnesses, etc.

[0068] The light source is a strip light source specially used for fill light of line scan cameras, and the brightness is adjustable.

[0069] The data collection process is described in detail below.

[0070] (1) Typical flat specimen

[0071] The size of the flat sample is 100mm×40mm×5mm. The upper surface of the sample is tested in an area of 100mm×40mm, with the upper left corner of the area as the origin.

[0072] Testing parameters (unit / mm):

[0073] (1) The integrated sensor (i.e., the integration of the line scan camera 2 and the spectral confocal sensor 3) moves to the position (X, Y, Z) = (0, 0, 50) and starts data acquisition;

[0074] (2) The three-axis motion platform drives the sensor to move along the X axis at a speed of 2 mm / s to 4 mm / s, moving to (100, 0, 50). The line scan camera acquisition frequency is 60 Hz to 100 Hz, and the profile data interval of the spectral confocal displacement sensor is 10 μm to 50 μm.

[0075] (3) Turn off the integrated sensor and move it to (X, Y, Z) = (0, 4, 50), start data acquisition, and move the sensor to (100, 4, 50);

[0076] (4) Repeat step (3), moving the sensor 4 mm in the Y direction each time. After 10 scans, a full inspection of the sample with a total width of 40 mm is completed.

[0077] (2) Typical cylindrical components

[0078] Typical dimensions of the component are: 30mm in diameter and 120mm in height. It is placed at the center of the electric rotary table. The cylindrical rotation axis is parallel to the Z-axis. The electric rotary table is 200mm away from the X-axis and the height of the electric rotary table is 200mm.

[0079] Testing parameters (unit / mm):

[0080] (1) The integrated sensor is rotated 90° counterclockwise and moved along the Y-axis until the sensor optical path intersects the centerline of the cylindrical component;

[0081] (2) Start the electric rotating table and adjust the rotation speed to 4-6 seconds per revolution;

[0082] (3) The sensor moves to (X, Y, Z) = (0, 200, 200), and the data acquisition function is turned on. The line scan camera acquisition frequency is 60 Hz to 100 Hz, and the profile data interval of the spectral confocal displacement sensor is 10 μm to 50 μm;

[0083] (4) After collecting data for one week according to the rotation speed of the electric rotary table, turn off the data collection, move 4 mm along the Z axis, and turn on the data collection function again;

[0084] (5) Repeat step (4) and scan 30 times to complete the scanning of the total height of the sample of 120 mm.

[0085] The following is a detailed description with reference to specific embodiments:

[0086] Before the test begins, the metal components to be tested must be prepared first. Taking the stainless steel metal flat plate component as an example, the size of the component is 100mm*40mm*5mm (length*width*height). After one year of exposure to the natural environment atmosphere, it has produced a lot of corrosion, and the accumulation of corrosion products has caused the surface to be uneven.

[0087] Cleaning and decontamination: Clean the surface of metal components by physical or chemical means (refer to GB / T 16545-1996 Corrosion of metals and alloys - Removal of corrosion products on corrosion test specimens) to ensure that corrosion products are completely removed and there is no external interference such as dust and oil that may affect corrosion detection.

[0088] Sample fixation: Fix the cleaned metal sample on an appropriate fixture to ensure the stability of the metal component during the test and avoid inaccurate test data due to sample movement.

[0089] The motion control platform synchronizes the data acquisition of both sensors, ensuring that the acquired image and depth information correspond. This ensures that color and depth data are precisely matched. Precise synchronization avoids data deviations caused by different acquisition times, achieving pixel-level spatiotemporal matching.

[0090] To detect color changes on metal surfaces caused by corrosion, the present invention uses a high-definition line scan camera to capture surface images. Through continuous line scan imaging, detailed color change data is obtained. Figure 3 This is a line scan corrosion detail image. Figure 3 As can be seen, corrosion on metal surfaces is often accompanied by color changes, with the appearance of corrosion features such as oxide layers and rust spots. These color changes can be extracted using image analysis algorithms to accurately locate the corroded areas and provide preliminary information on the corrosion.

[0091] Comprehensively assessing the extent of corrosion on a metal surface requires obtaining depth information within the corroded area. Using a spectral confocal displacement sensor, you can accurately measure the three-dimensional topography of a metal surface. Spectral confocal technology precisely measures surface height variations, particularly near corrosion points, enabling you to determine the depth of corrosion. This improves measurement accuracy by eliminating interference from stray light from other surfaces. Furthermore, the spectral confocal displacement sensor boasts micron-level depth measurement accuracy, enabling you to record depth information for each corrosion point, helping you accurately assess the severity of corrosion damage.

[0092] Figure 4 It is a spectral confocal displacement sensor that collects data. You can choose different data intervals to export test data, export CSV data in units of 0.8mm, and select any data to create the following Figure 4 In the line chart. Figure 4 In the table, the rows represent the position coordinates of the sample in the X direction, and the columns represent the position coordinates of the sample in the Y direction. The position difference between two adjacent tables is the data interval exported from CSV data (0.8mm in the example). The values in the table represent the depth data of the points corresponding to the row and column coordinates. Figure 4 The broken line in the line graph represents the depth change information corresponding to the row of data.

[0093] Based on this line graph, the maximum depth and average depth of the area can be obtained.

[0094] The metal surface morphology is modeled through the collected data and a three-dimensional image is generated. Figure 5 shown.

[0095] One aspect of the present invention provides a composite optical detection method for metal surface corrosion, comprising:

[0096] Through the spatiotemporal registration method, a line scan camera and a spectral confocal sensor are used for synchronous data acquisition. The line scan camera is used to obtain RGB image data (i.e., color data) of the metal surface to be inspected, and the spectral confocal sensor is used to obtain point cloud data (i.e., topographic data) of the metal surface to be inspected.

[0097] Perform multimodal data fusion on RGB data and point cloud data;

[0098] By introducing confidence weights, the data fusion results are optimized;

[0099] A dual-threshold segmentation method of color threshold and depth threshold is used to preliminarily identify the corrosion area in the optimized data fusion results;

[0100] Through opening and closing operations, the preliminary recognition results are subjected to morphological post-processing to generate the final recognition results;

[0101] The final identification results are used to quantitatively evaluate the corrosion area.

[0102] Through the data synchronization interface, the high-definition line scan camera and the spectral confocal displacement sensor realize real-time synchronous data transmission through the data bus, ensuring the consistency of the data timestamps of the two, which facilitates the subsequent data fusion processing. The spatiotemporal registration method includes:

[0103] Determine the axis distance and scanning speed of the line scan camera and spectral confocal sensor;

[0104] Determine the time delay based on the wheelbase and scanning speed;

[0105] The trigger timing of the line scan camera and the spectral confocal sensor is adjusted according to the time delay, thereby ensuring that the timestamps of the data collected by the line scan camera and the spectral confocal sensor are aligned.

[0106] For example, in the process of synchronous corrosion data acquisition, the fixed axis distance between the line scan camera and the spectral confocal displacement sensor is set to ΔL, and the scanning speed is v. The time delay Δt is:

[0107]

[0108] By adjusting the sensor trigger timing, the color (RGB) and three-dimensional coordinate (XYZ) data timestamps of the same spatial point are aligned.

[0109] After the data is acquired, it can be processed. Figure 6 is the data fusion flow chart. Figure 6 As shown in the figure, the collected data needs to be spatiotemporally registered (by calculating time delay and data alignment), and then undergo data fusion (including six-dimensional data generation, confidence weight assignment and feature-level enhancement) and output the fusion result.

[0110] By introducing confidence weights, the optimized data fusion results include:

[0111] According to the weight adjustment strategy, the color weight of the RGB image data and the morphological weight of the point cloud data (i.e., the point cloud data after data fusion) are adaptively adjusted. The weight adjustment strategy includes: determining the corresponding weight according to the noise level of the line scan camera and the spectral confocal sensor, adjusting the weight according to the corrosion depth, adjusting the weight according to the image contrast, etc.

[0112] For example, for each measurement point, the fused six-dimensional dataset D can be expressed as:

[0113] D = (X, Y, Z, R, G, B)

[0114] Among them, (X, Y, Z) is the three-dimensional shape data, and (R, G, B) is the color information. Figure 5The three-dimensional shape data in Figure 7 is the data fusion result diagram, where Figure 7 The middle (a) figure shows the image obtained by line scanning. Figure 7 The middle (b) figure is the fused data.

[0115] Next, the confidence weights a (shape weight) and b (color weight) are introduced to optimize the fusion results and generate optimized fusion data D fused :

[0116] D fused =a·(X,Y,Z)+b·(R,G,B).

[0117] The weight value is dynamically adjusted according to the sensor accuracy, for example: Among them, σ z and σ RGB Represent the standard deviation of the topographic data and color data respectively.

[0118] For example, spectral confocal sensors have low noise (σ z =1μm), while the color noise caused by illumination changes is high (σ RGB =10), the weights are: a=1, b=0.01. At this time, the morphology data dominates the fusion result.

[0119] For areas with high corrosion depth, the morphology weight a can be increased and the color weight b can be reduced; for areas with low contrast: the color weight b can be increased to supplement the missing morphology data.

[0120] For the edge area of the corrosion area, enhanced fusion can also be performed through adaptive feature-level fusion. That is, based on the corrosion feature extraction, the edge area of the corrosion area is enhanced and fused.

[0121] The enhanced fusion of key areas (such as the edge of the corrosion pit) can be expressed as:

[0122] D enhanced =r·D fused +(1-r)·D edge ,

[0123] Among them, r represents the global fusion coefficient, which is usually set to 0.6~0.75, D edge Represents the data of the corrosion edge area, D enhanced Represents the result of enhanced fusion, by which edge details can be highlighted.

[0124] The influence of low-quality data is suppressed by the confidence weight of feature fusion, and the anti-noise ability is enhanced; the enhanced fusion based on feature extraction prioritizes the retention of key information such as the edge of the corrosion pit; moreover, the fusion strategy is automatically adjusted according to the characteristics of the corrosion area, which can be applied to complex surfaces (such as curved surfaces, special-shaped components, etc.).

[0125] After data fusion, the corrosion areas in the optimized data fusion results can be preliminarily identified, which includes the following steps:

[0126] Convert the RGB image data in the optimized data fusion result into HSV image data, and extract the H channel, S channel and V channel information;

[0127] Set the color threshold of H channel, S channel and V channel;

[0128] According to the point cloud data in the optimized data fusion result, the depth information is normalized to generate the reference plane Z base ;

[0129] Using the reference plane Z base , generating the corrosion depth threshold;

[0130] Based on the color threshold and corrosion depth threshold, the preliminary identification results of the corrosion area are generated.

[0131] Specifically, the RGB image is converted to HSV space, and the hue channel (i.e., H channel) is used to separate the corrosion color (such as the reddish-brown color of rust). The depth data Z obtained by the spectral confocal displacement sensor is normalized to eliminate the influence of sample tilt or installation error.

[0132] After statistical analysis, in the HSV space, the hue range of the corrosion area is mainly concentrated in 0°~30° and 330°~360°. Therefore, the hue channel color threshold can be set to 0°~30° and 330°~360°, and the color threshold of the S channel can be set to T S (For example, it can be set to 50) and the color threshold T of the V channel V (For example, it can also be set to 50). If the hue range is no longer 0°~30° and 330°~360°, or the saturation is low (ie S<T S ), or low brightness (ie V<T V ), these interference areas can be removed.

[0133] In addition, you can also use the reference plane to set the depth threshold. Select the depth mean Z of the uncorroded area base As a benchmark (i.e., reference surface), the corrosion depth threshold T Z Set to: T Z =Z base-ΔZ, where ΔZ is the minimum depth of the corrosion pit, which can usually be set to 10μm.

[0134] Then, the double threshold segmentation method is used to generate the preliminary identification results of the corrosion area.

[0135] Corrosion area = {(H, S, V, Z) | H∈T H ∧S≥T S ∧V≥T V ^Z≤T Z},

[0136] Among them, T H =0°~30°∪330°~360°, ^ indicates that the conditions need to be met at the same time (i.e., logical "AND" operation).

[0137] The double threshold segmentation function is:

[0138]

[0139] Through the double threshold function, the positions belonging to the corrosion area are marked as 1, and the other positions are marked as 0, and a binary mask is generated to obtain the preliminary identification result of the corrosion area.

[0140] The dual-threshold processing method described above enhances the ability to resist light interference, as color segmentation alone is susceptible to uneven lighting or luminescence, leading to false detections. After introducing dual-threshold segmentation, depth information eliminates interference from non-depressed areas, significantly reducing the false detection rate. Furthermore, the dual-threshold method can improve the detection rate of shallow corrosion, as depth segmentation alone is difficult to detect incipient corrosion with minimal depth changes (e.g., corrosion depth <20μm). Combining this with color changes (e.g., hue shifts caused by corrosion) can effectively improve the detection rate. The dual-threshold method also enhances its noise resistance, as if noise is present in the line scan camera or spectral confocal sensor, the noise points can be suppressed through a logical "AND" operation, improving the accuracy of corrosion damage segmentation.

[0141] Then, the preliminary recognition results are morphologically post-processed through opening and closing operations to generate the final recognition results.

[0142] Morphological processing is a local processing of an image based on a structural element. The structural element is a small binary matrix, usually square, circular or cross-shaped. Two basic morphological operations include erosion and dilation.

[0143] The erosion operation traverses the image A by sliding the structural element B, retaining the area completely covered by the structural element. Its formula can be expressed as:

[0144]

[0145] Among them, (B)z Represents the position of the structural element B after translation by z.

[0146] The dilation operation traverses the image A by sliding the structural element B, retaining the area where the structural element intersects the image. The formula is as follows:

[0147]

[0148] The binary mask is opened (to remove noise) and closed (to fill holes) to optimize the eroded area.

[0149] For the binary image I, an opening operation is performed, that is, an erosion operation is performed first, and then an expansion operation is performed. The erosion operation can be expressed as:

[0150]

[0151] The expansion operation can be expressed as:

[0152] For the closing operation, that is, the expansion operation is performed first, and then the erosion operation is performed, then the closing operation for the binary image I after the opening operation can be expressed as:

[0153]

[0154] Among them, I dilated Represents the result of the expansion operation in the closing operation, I closed Represents the result of the corrosion operation in the closing operation, that is, the final result of the closing operation.

[0155] The following describes this with specific examples.

[0156] (1) Input image

[0157] Assume that the input binary image (i.e., binary mask) I, where the eroded area is the foreground (value 1) and the non-eroded area is the background (value 0). Assume that I can be expressed as:

[0158]

[0159] The structural element B is a 3×3 square matrix:

[0160]

[0161] (2) Open operation to eliminate noise

[0162] Perform erosion operation on image I:

[0163]

[0164] Perform dilation on the eroded image:

[0165]

[0166] Image I after opening operation opened Small noise and isolated points are eliminated.

[0167] (3) Closing operation to fill holes

[0168] Image I after split operation opened Perform the expansion operation:

[0169]

[0170] Perform erosion on the expanded image:

[0171]

[0172] Image I after closing operation closed Small holes are filled and broken areas are connected.

[0173] Then, the final optimized binary image I final =I closed Morphological post-processing can effectively eliminate noise, fill holes, and optimize the boundaries of eroded areas through a combination of opening and closing operations.

[0174] Next, the final identification results are used to quantitatively assess the corrosion area. This process may include:

[0175] Based on the final identification results, determine the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature;

[0176] The corrosion index is determined by taking a weighted sum of the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature;

[0177] Use the corrosion index to determine the corrosion level.

[0178] Corrosion area ratio A corrosion It can be expressed as:

[0179]

[0180] Maximum corrosion depth Z max It can be expressed as:

[0181] Z max =max(Z base -Z corrosion )

[0182] Average corrosion depth Z avg It can be expressed as:

[0183]

[0184] where Z base Z is the reference height of the uncorroded area. corrosion represents the depth of each real corrosion pit, and N represents the number of corrosion pits.

[0185] Curvature is a geometric quantity that describes the degree of curvature of a curve or surface. For corrosion 3D point cloud data collected by a spectral confocal sensor, curvature can be calculated using the second-order derivative of the local surface. Specifically, the determination of corrosion morphology curvature includes:

[0186] Obtain the point cloud data of the erosion area segmented after data fusion (i.e., the final recognition result obtained by the optimized data fusion result);

[0187] Calculate the second derivative of the depth data for each point in the corrosion area;

[0188] The absolute values of the second derivatives of the depth data at each point in the corrosion area are averaged;

[0189] The average value is taken as the corrosion morphology curvature of the corrosion area.

[0190] The curvature C of the corrosion pit is defined as the average value of the local surface curvature, that is:

[0191]

[0192] where Z j is the depth value of the jth point in a corrosion area, and M represents the total number of points in a corrosion area;

[0193] is the second-order derivative of the depth data (Laplacian operator), which represents the curvature of the local surface.

[0194] For example, after acquiring the point cloud data, Gaussian filtering can be used to denoise the three-dimensional point cloud data acquired by the spectral confocal displacement sensor, and then the point cloud data can be converted into a regular triangular mesh to facilitate the calculation of curvature.

[0195] For each point j, fit a local quadratic surface using its neighborhood points (3×3 grid):

[0196] Z(x,y)=ax 2 +by 2 +cxy+dx+ey+f

[0197] Where a, b, c, d, e, and f are fitting coefficients.

[0198] Compute the curvature using the coefficients of the fitted surface:

[0199]

[0200] Take the absolute value of the curvature at all points in the eroded region and average it:

[0201]

[0202] High curvature area (C>0.5): The corrosion pit has sharp edges and complex geometric shapes; it is easy to cause stress concentration and lead to crack propagation; it is common in localized corrosion types such as pitting corrosion.

[0203] Low curvature area (C<0.3): The surface of the corrosion pit is flat and the geometry is simple; it has little effect on the structural strength; it is common in uniform corrosion and slight oxidation.

[0204] Based on the aforementioned multi-dimensional corrosion data, the present invention designs a corrosion severity index (CSI), namely, a corrosion index.

[0205] In one embodiment, the corrosion index can be expressed as:

[0206]

[0207] Among them, A represents the percentage of corrosion area (%), A max is the maximum allowable corrosion area (e.g. 35%);

[0208] Z max Indicates the maximum corrosion depth (μm), Z critical is the critical depth (e.g. 10% of the material thickness);

[0209] Z avg Indicates the average corrosion depth (μm);

[0210] C represents the curvature of the corrosion area (dimensionless), C critical is the critical curvature (e.g. 0.5%);

[0211] α, β, γ, and δ represent the set weight coefficients respectively. The weight coefficients can be flexibly adjusted according to the material type or application scenario. For example, for structural steel, the weight can be set to:

[0212] α=0.2, β=0.4, γ=0.3, δ=0.1

[0213] For different metal materials in different application scenarios, the weight coefficient adjustment method is shown in Table 1.

[0214] Table 1 Weight setting methods for different scenarios

[0215]

[0216] By calculating the corrosion index, the corrosion degree can be classified into different levels. For example, the corrosion level can be divided into two or more levels by setting different thresholds. Table 2 is an exemplary corrosion degree classification result.

[0217] Table 2 Corrosion degree classification

[0218] CSI Range Corrosion level Severity Description CSI<0.3 Level 1 Mild corrosion: localized small-scale corrosion with no significant impact on structural performance 0.3≤CSI<0.7 Level 2 Moderate corrosion: The corrosion area or depth increases significantly, which may affect the strength CSI ≥ 0.7 Level 3 Severe corrosion: Large-scale or deep corrosion, with the risk of structural failure

[0219] In some embodiments, the final identification results can also be used to quantitatively assess the corrosion area, including a combination of one or more of the following: corrosion area percentage, maximum corrosion depth, average corrosion depth, and corrosion morphology curvature, to determine the corrosion type. Based on the results of different characteristic parameters, the corrosion type can be classified into uniform corrosion, pitting corrosion, general corrosion, etc. Table 3 shows an exemplary classification result.

[0220] Table 3 Corrosion type classification

[0221]

[0222] Figure 8 Given a corroded structural steel. After testing and analysis, the corrosion area accounts for A = 27.8%, and the maximum corrosion depth Z max =83μm, average corrosion depth Z avg =32μm, curvature C=0.65.

[0223] Set threshold: A max =30%, Z critical =100μm, C critical =0.5.

[0224] Then, the corrosion severity index CSI can be calculated:

[0225]

[0226] Since CSI=0.72>0.7, it can be classified as level 3 severe corrosion.

[0227] Combined with various characteristic parameter combinations, that is, A<50%, Z max >0.8Z critical , C>0.5, therefore, it can be determined that the corrosion type is pitting corrosion.

[0228] In summary, the multimodal sensing collaboration adopted in the present invention, that is, the use of different sensors to realize the synchronous acquisition of "image-morphology" based on a fixed wheelbase optical path, solves the problem of traditional multi-sensor data misalignment and ensures the spatial consistency of corrosion characteristics; moreover, by fusing multi-dimensional information, combining high-definition line scan cameras and spectral confocal displacement sensors, the synchronous acquisition and fusion of metal surface color and size information are realized, and the corrosion state of the metal surface is more comprehensive and accurate; embedded real-time data processing can be deployed on edge devices to realize rapid on-site detection with low dependence on computing power; the proposed feature-level fusion strategy realizes pixel-level data fusion characterization.

[0229] Therefore, the hardware and software of the device of the present invention together constitute an efficient automated corrosion detection platform. The system integrates modules such as a high-definition line scan camera, a spectral confocal displacement sensor, a motion control platform, and a data fusion processing unit. It can automatically complete the detection and evaluation of surface corrosion of metal components based on the set detection parameters.

[0230] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A composite optical detection method for metal surface corrosion, characterized in that: include: Through the spatiotemporal registration method, a line scan camera and a spectral confocal sensor are used for synchronous data acquisition. The line scan camera is used to obtain RGB image data (i.e., color data) of the metal surface to be inspected, and the spectral confocal sensor is used to obtain point cloud data (i.e., topographic data) of the metal surface to be inspected. Perform multimodal data fusion on RGB data and point cloud data; By introducing confidence weights, the data fusion results are optimized; A dual-threshold segmentation method of color threshold and depth threshold is used to preliminarily identify the corrosion area in the optimized data fusion results; Through opening and closing operations, the preliminary recognition results are subjected to morphological post-processing to generate the final recognition results; The final identification results are used to quantitatively evaluate the corrosion area.

2. The composite optical detection method for metal surface corrosion according to claim 1, characterized in that: Spatiotemporal registration methods include: Determine the axis distance and scanning speed of the line scan camera and spectral confocal sensor; Determine the time delay based on the wheelbase and scanning speed; The trigger timing of the line scan camera and the spectral confocal sensor is adjusted according to the time delay, thereby ensuring that the timestamps of the data collected by the line scan camera and the spectral confocal sensor are aligned.

3. The composite optical detection method for metal surface corrosion according to claim 1, characterized in that: By introducing confidence weights, the data fusion results are optimized, including: According to the weight adjustment strategy, the color weight of the RGB image data and the morphology weight of the point cloud data are adaptively adjusted. The weight adjustment strategy includes: determining the corresponding weight according to the noise level of the line scan camera and the spectral confocal sensor, adjusting the weight according to the corrosion depth, adjusting the weight according to the image contrast, etc.

4. The composite optical detection method for metal surface corrosion according to claim 1, characterized in that: The method further comprises performing enhanced fusion on the edge area of the corrosion area based on corrosion feature extraction.

5. The composite optical detection method for metal surface corrosion according to claim 1, characterized in that: The initial identification of the corrosion area in the optimized data fusion results includes: Convert the RGB image data in the optimized data fusion result into HSV image data, and extract the H channel, S channel and V channel information; Set the color threshold of H channel, S channel and V channel; According to the point cloud data in the optimized data fusion result, the depth information is normalized to generate the reference surface ; Using a datum , generating the corrosion depth threshold; Based on the color threshold and corrosion depth threshold, the preliminary identification results of the corrosion area are generated.

6. The composite optical detection method for metal surface corrosion according to claim 1, characterized in that: The final identification results are used to quantitatively evaluate the corrosion area, including: Based on the final identification results, determine the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature; The corrosion index is determined by taking a weighted sum of the corrosion area ratio, maximum corrosion depth, average corrosion depth and corrosion morphology curvature; Use the corrosion index to determine the corrosion level.

7. The composite optical detection method for metal surface corrosion according to claim 6, characterized in that: The determination of the curvature of the corrosion morphology includes: Obtain the point cloud data of the corrosion area segmented after data fusion; Calculate the second derivative of the depth data for each point in the corrosion area; The absolute values of the second derivatives of the depth data at each point in the corrosion area are averaged; The average value is taken as the corrosion morphology curvature of the corrosion area.

8. The composite optical detection method for metal surface corrosion according to claim 6, characterized in that: Using the final identification results, a quantitative assessment of the corrosion area also includes a combination of one or more of the corrosion area ratio, maximum corrosion depth, average corrosion depth, and corrosion morphology curvature to determine the corrosion type.

9. A composite optical detection system for metal surface corrosion, characterized in that: include: Three-axis motion control platform, line scan camera, spectral confocal sensor and data processing equipment, The three-axis motion control platform includes three mutually perpendicular motion axes; The line scan camera and the spectral confocal sensor are connected to the same motion axis of the three-axis motion control platform, and the wheelbase between the line scan camera and the spectral confocal sensor is fixed. The line scan camera and the spectral confocal sensor are controlled to move along the three motion axes by a servo motor. The line scan camera is used to obtain RGB image data, i.e., color data, of the metal surface to be inspected; The spectral confocal sensor is used to obtain point cloud data, i.e., topographic data, of the metal surface to be inspected; The data processing device is used to execute the composite optical detection method for metal surface corrosion according to any one of claims 1 to 8.

10. The composite optical detection system for metal surface corrosion according to claim 9, characterized in that: The system further includes a motorized rotating stage, and the integrated sensor of the line scan camera and the spectral confocal sensor can be rotated by servo motor control.

Images