Ice aggregate grading intelligent detection method based on image recognition
By combining ultraviolet channel images and structured light reflectance images, the problem of real-time detection of ice aggregate gradation and frozen water content was solved, enabling accurate detection of ice aggregate gradation and frozen water content, providing reliable data support, and ensuring stable quality of the mixture.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUZHOU AUTOMOBILE RES INST OF TSINGHUA UNIV (WUJIANG)
- Filing Date
- 2026-01-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies cannot accurately detect the gradation of ice aggregates and the content of frozen water in real time, leading to uncontrolled slump and early strength fluctuations in the mixture, and failing to provide reliable data to support real-time proportioning adjustments.
A method combining ultraviolet channel images and structured light reflection images is used to achieve real-time detection of ice aggregate gradation and frozen water content through signal extraction, fringe frequency analysis, and coupling intensity distribution mapping. This includes ultraviolet light excitation of cold light radiation, structured light projection, image acquisition, signal processing, and parameter calculation.
It enables real-time and accurate detection of ice aggregate gradation and frozen water content, breaking through the limitations of traditional detection methods, providing reliable data support, and providing closed-loop regulation for the quality control of the mixture.
Smart Images
Figure CN122023906A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image recognition technology, and in particular to an intelligent detection method for aggregate gradation based on image recognition. Background Technology
[0002] In engineering sites with high altitudes or seasonally low temperatures, such as high-speed rail beam fabrication yards, airport runway repair facilities, and precast concrete workshops at polar research stations, crushed stone aggregates are often stored in environments below 0°C or rapidly cooled by artificial spraying, inevitably resulting in the formation of uneven ice films and frozen water on their surfaces. When these icy aggregates are transported to the mixing system, their particle size distribution, shape parameters, and moisture content dynamically change depending on the thickness of the ice film.
[0003] Current testing methods mainly rely on manual sieving, weighing, or sampling with a single optical sensor. These methods cannot penetrate the semi-transparent ice film to identify the true aggregate outline and lack the ability to quantify the frozen water content in real time. On high-speed production lines, this kind of lag-based testing is difficult to provide reliable data support for real-time proportioning adjustments and can easily lead to quality problems such as uncontrolled slump of the mixture, early strength fluctuations, and surface bleeding. Summary of the Invention
[0004] This invention provides an intelligent detection method for aggregate gradation based on image recognition. Its main purpose is to solve the problem that the existing hysteresis estimation detection method is difficult to provide reliable data support for real-time proportion adjustment.
[0005] To achieve the above objectives, the present invention provides an intelligent detection method for aggregate gradation based on image recognition, comprising: S1. Collect ultraviolet channel images and structured light reflection images of the ice aggregate detection area; S2. Extract signals from the ultraviolet channel image to obtain the TCP signal feature image, and perform stripe frequency analysis on the structured light reflection image to obtain the PDR stripe frequency distribution map. S3. Perform pixel-by-pixel weighted fusion calculation on the TCP signal feature image and the PDR stripe frequency distribution map to obtain the coupling strength distribution map; S4. Based on the statistical parameters of the coupling strength distribution map, the ice film area of the ice aggregate detection area is binarized to obtain the ice film mask; S5. Generate aggregate candidate regions based on ultraviolet channel images and structured light reflection images, and calculate the actual size parameters of ice aggregates in the ice aggregate detection region based on the aggregate candidate regions and ice film mask. S6. Calculate the freezing water correction factor for the average coupling intensity of the coupling intensity distribution map to obtain the freezing water content in the ice aggregate detection area; S7. Determine whether the ice aggregate is qualified based on the frozen water content and actual size parameters.
[0006] In a preferred embodiment, acquiring ultraviolet channel images and structured light reflectance images of the ice aggregate detection area includes: In response to the detection signal At any given moment, the TCP material in the ice aggregate detection area is excited by an ultraviolet pulse light source to generate cold light radiation, and ultraviolet channel images of the ice aggregate detection area are acquired. exist At ms, striped structured light is projected onto the ice aggregate through a striped structured light projector, and the structured light reflection image of the ice aggregate detection area is acquired.
[0007] In a preferred embodiment, signal extraction is performed on the ultraviolet channel image to obtain a TCP signal feature image, and fringe frequency analysis is performed on the structured light reflection image to obtain a PDR fringe frequency distribution map, including: The background image of the area without aggregate is acquired, and pixel-by-pixel difference operation is performed on the ultraviolet channel image and the background image to obtain the TCP raw difference image; The TCP raw differential image is subjected to noise cancellation processing to obtain the TCP signal feature image; A two-dimensional Fourier transform is performed on the structured light reflection image to obtain the frequency domain amplitude spectrum of the structured light reflection image; The energy peak coordinates of the frequency domain amplitude spectrum are retrieved using a peak detection algorithm. Calculate the frequency value of the energy peak coordinate point; by Centered on, radius Define the frequency domain region, extract the frequency components of the frequency domain region, and perform inverse Fourier transform on the frequency components to obtain the reconstructed fringe pattern. The absolute phase value of each pixel in the reconstructed striped pattern is calculated using a spatial phase analysis algorithm. Substitute the absolute phase value into the calibration formula to calculate the PDR stripe frequency value for each pixel. A PDR fringe frequency distribution map is generated based on the PDR fringe frequency values.
[0008] In a preferred embodiment, a pixel-by-pixel weighted fusion calculation is performed on the TCP signal feature image and the PDR stripe frequency distribution map to obtain a coupling strength distribution map, including: After spatially registering the TCP signal feature image and the PDR stripe frequency distribution map, extract the TCP cold light intensity value and PDR stripe frequency value of the pixels at the same position. The coupling strength value is obtained by weighting the TCP cold light intensity value and the PDR fringe frequency value. Generate a coupling strength distribution map based on the coupling strength value.
[0009] In a preferred embodiment, the ice film region of the ice aggregate detection area is binarized based on statistical parameters of the coupling intensity distribution map to obtain an ice film mask, including: The threshold for determining ice film is calculated based on statistical parameters; If the coupling strength value at a pixel in the coupling strength distribution map is greater than the ice film determination threshold, the pixel is marked as an ice film region and assigned a value of 1; otherwise, the pixel is marked as a non-ice film region and assigned a value of 0. An ice mask is generated based on each pixel of the coupling strength distribution map.
[0010] In a preferred embodiment, generating aggregate candidate regions based on the ultraviolet channel image and the structured light reflectance image includes: Grayscale correction and Gaussian filtering noise reduction are performed on ultraviolet channel images and structured light reflection images; The pre-screening region of aggregates is extracted by edge detection, and candidate regions of aggregates are selected from the pre-screening region by connected component analysis.
[0011] In a preferred embodiment, the actual size parameters of the ice aggregate in the ice aggregate detection area are calculated based on the aggregate candidate area and the ice film mask, including: Morphological treatment of the ice film mask; Under the constraint of the aggregate candidate region, the contour of the morphologically processed ice film mask is extracted to obtain the real aggregate contour image; The actual size parameters of the ice aggregate are calculated based on the actual aggregate outline image. The actual size parameters include the actual perimeter, actual length, actual width, actual area, actual aspect ratio, and actual shape factor.
[0012] In a preferred embodiment, the average coupling intensity of the coupling intensity distribution map is converted using a freezing water correction factor to obtain the freezing water content of the ice aggregate detection area, including: Calculate the average coupling strength of the coupling strength distribution map; Substituting the average coupling strength into the modified formula yields the frozen water content.
[0013] In a preferred embodiment, the calculation formula for the correction formula is as follows:
[0014] In the formula, It is the frozen water content. It is the frozen water correction factor. It is the average coupling strength.
[0015] In a preferred embodiment, determining whether the ice aggregate is qualified based on the frozen water content and actual size parameters includes: The actual size parameters of all ice aggregates were grouped according to particle size range to obtain multiple particle size distribution groups; The number of ice aggregates within each particle size distribution group and the total number of ice aggregates in the ice aggregate testing area are statistically analyzed. The ratio of the number of ice aggregates within each group to the total number of ice aggregates is calculated to obtain the proportion of ice aggregates in each particle size distribution group. Calculate the total area of all ice aggregates within the particle size distribution group based on the actual area, and calculate the ratio of the total area within the group to the total area of ice aggregates in the ice aggregate detection area to obtain the area ratio of ice aggregates in the particle size distribution group. For each particle size distribution group, the shape category of ice aggregate is determined based on the actual aspect ratio and actual shape factor. The proportion of ice aggregate quantity under each shape category is counted. The ratio of the proportion of ice aggregate quantity under each category to the total number of ice aggregates in the group is calculated to obtain the proportion of ice aggregate shape category in the particle size distribution group. The quality of ice aggregate is determined based on the proportion of aggregate quantity, ice aggregate area, ice aggregate shape category, and frozen water content.
[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention first arranges two sets of active light fields on the ice aggregate conveyor belt: an ultraviolet pulse light source and a striped structured light projector. The ultraviolet channel image and the structured light reflection image are acquired in the same detection window. The ultraviolet excitation can cause the TCP material in the aggregate to generate cold light radiation, revealing the real mineral boundaries hidden under the ice film. The structured light stripes are extremely sensitive to surface geometric undulations and can record the stripe distortion caused by the difference in ice film thickness. The two are triggered synchronously in time and spatially registered at the pixel level, providing a complementary spectral-morphological information basis for subsequent algorithms, breaking through the limitation of traditional single optical or manual screening that cannot penetrate the semi-transparent ice film.
[0017] 2. This invention first performs pixel-level difference analysis between the ultraviolet channel image and the background image, and then uses adaptive filtering to suppress interference and enhance TCP cold light characteristics. Next, it performs two-dimensional Fourier transform and phase expansion on the structured light reflection image to obtain the PDR stripe frequency. Subsequently, it fuses the cold light intensity of the same pixel with the stripe frequency according to weight to form a coupling intensity map. The ice film is then quickly segmented using a statistical threshold. Combined with aggregate candidate regions and morphological optimization, the system obtains the real aggregate outline and calculates parameters such as perimeter, length and width, area, aspect ratio, and shape factor in real time. This enables dimensional correction and online reconstruction of the ice film magnification effect, providing reliable data support for closed-loop control of aggregate gradation and frozen water content.
[0018] 3. This invention introduces a freezing water correction coefficient into the average value of the coupling strength distribution map for linear mapping, and calculates the freezing water content of the detection area in real time. Then, it statistically analyzes the proportion of aggregate quantity, area, and shape category according to particle size range, and constructs a multi-dimensional evaluation matrix together with the freezing water content. When all indicators fall within the preset qualified threshold, the aggregate is judged to be qualified; otherwise, an alarm is immediately triggered and feedback is sent to the mixing control unit, realizing closed-loop adjustment of process parameters such as slump and strength. It solves the core technical problem of synchronous online detection of particle size distribution and freezing water content under ice film obstruction in a coordinated manner at the three levels of sensing, algorithm, and decision-making. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating an intelligent detection method for ice aggregate gradation based on image recognition, provided in an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0020] Reference Figure 1 The diagram shown is a flowchart illustrating an intelligent detection method for ice aggregate gradation based on image recognition, according to an embodiment of the present invention. In this embodiment, the intelligent detection method for ice aggregate gradation based on image recognition includes: 1. A method for intelligent detection of aggregate gradation based on image recognition, characterized in that the method comprises: S1. Collect ultraviolet channel images and structured light reflection images of the ice aggregate detection area; In an embodiment of the present invention, acquiring ultraviolet channel images and structured light reflectance images of the ice aggregate detection area includes: In response to the detection signal At any given moment, the TCP material in the ice aggregate detection area is excited by an ultraviolet pulse light source to generate cold light radiation, and ultraviolet channel images of the ice aggregate detection area are acquired. exist At ms, striped structured light is projected onto the ice aggregate through a striped structured light projector, and the structured light reflection image of the ice aggregate detection area is acquired.
[0021] Specifically, the detection signal refers to the instruction signal that triggers the ice aggregate detection process; The "moment" refers to the initial moment after the detection signal is triggered; "TCP material" refers to the cold light emitting material contained in the ice aggregate, such as specific fluorescent agents and mineral impurities, which release cold light after being excited by ultraviolet light; "structured light reflection image" refers to the projection of a specific pattern of structured light onto the detection object using a striped structured light projector. When the structured light shines on the surface of the ice aggregate, it is reflected and deformed due to factors such as the surface morphology and physical properties of the ice aggregate, and then the image is captured and recorded by the image acquisition device. Its pixel information can reflect the morphological changes after the interaction between the structured light and the surface of the ice aggregate, and can carry the geometric morphological features of the ice aggregate surface, such as the surface concavity, contour, size, and other information; "ultraviolet channel image" is an image obtained by exciting the detection area of the ice aggregate with an ultraviolet pulse light source, causing specific substances in the ice aggregate to produce light radiation, and then being acquired by the image acquisition device in the ultraviolet light band. This image can capture the optical response characteristics of the ice aggregate under ultraviolet excitation conditions and can reflect the material composition, distribution state, and other material-related information of the ice aggregate.
[0022] In detail, when the detection signal is at time Upon arrival, a controlled ultraviolet pulse light source is used to irradiate the ice aggregate detection area. The ultraviolet pulse light source excites the TCP material in the ice aggregate within the detection area, causing it to produce cold light radiation. At this time, an image acquisition device captures an image of the detection area, obtaining an ultraviolet channel image. This image records the cold light radiation characteristics of the ice aggregate under ultraviolet light excitation. Then, at time... At a certain time (ms), the striped structured light projector is activated to project striped structured light onto the ice aggregate detection area. The striped structured light will be reflected and deformed on the ice aggregate surface due to differences in surface morphology. At this time, the detection area is photographed again using an image acquisition device to obtain a structured light reflection image. This image contains information about the deformation of the structured light stripes caused by the surface morphology of the ice aggregate. By acquiring ultraviolet channel images and structured light reflection images at different times, image data reflecting different characteristics of the ice aggregate can be obtained, providing a foundation for subsequent analysis and processing.
[0023] S2. Extract signals from the ultraviolet channel image to obtain the TCP signal feature image, and perform stripe frequency analysis on the structured light reflection image to obtain the PDR stripe frequency distribution map. In an embodiment of the present invention, signal extraction is performed on the ultraviolet channel image to obtain a TCP signal feature image, and stripe frequency analysis is performed on the structured light reflection image to obtain a PDR stripe frequency distribution map, including: The background image of the area without aggregate is acquired, and pixel-by-pixel difference operation is performed on the ultraviolet channel image and the background image to obtain the TCP raw difference image; In detail, the TCP raw differential image is an image data carrier obtained after specific image processing operations. It is used to highlight the TCP cold light radiation related characteristics in the ice aggregate detection scenario, effectively counteract the inherent interference of the environment, and accurately present the cold light radiation signal generated by the TCP material in the ice aggregate.
[0024] In detail, a background image of the aggregate-free detection area is pre-acquired. This background image must be acquired under the same environmental conditions as the UV channel image to ensure consistency of background interference. After the background image acquisition is completed, for each pixel at corresponding positions in the two images, the pixel grayscale value of the UV channel image is subtracted from the pixel grayscale value of the background image. By traversing all pixels, the original TCP differential image is finally obtained, thereby eliminating the interference of the environmental background on the cold light radiation signal of the ice aggregate TCP and highlighting the cold light radiation characteristics generated by the ice aggregate TCP material.
[0025] The TCP raw differential image is subjected to noise cancellation processing to obtain the TCP signal feature image; In detail, the TCP signal feature image is an image product obtained after noise cancellation processing of the original TCP differential image. It eliminates irrelevant signals such as random environmental interference and equipment acquisition noise in the original TCP differential image, retains and enhances the effective features generated by the TCP cold light radiation of ice aggregate, and enables the image to accurately present key information such as the spatial distribution and intensity changes of the TCP signal.
[0026] In detail, an adaptive median filtering algorithm is used to cancel noise in the TCP raw differential image. Each pixel of the TCP raw differential image is traversed, and an initial filtering window is determined with the current pixel as the center. The window size is dynamically adjusted based on the local noise density of the image. The noise density is determined by comparing the standard deviation of the pixel grayscale values within the window with a preset threshold. For each pixel within the window, the median grayscale value is calculated. If the deviation between the current center pixel's grayscale value and the median is greater than a noise threshold determined based on the overall grayscale distribution of the image, the pixel is identified as a noise pixel, and its grayscale value is replaced with the median. If the deviation is less than or equal to the threshold, the center pixel's grayscale value is retained.
[0027] A two-dimensional Fourier transform is performed on the structured light reflection image to obtain the frequency domain amplitude spectrum of the structured light reflection image; In detail, the pixel grayscale values of the structured light reflection image are treated as two-dimensional discrete signals. Based on the mathematical definition of the two-dimensional Fourier transform, the frequency domain representation of this discrete signal is calculated. Specifically, for the grayscale value corresponding to each pixel position in the image, a double summation operation is performed to convert it into a frequency domain function with frequency variables as coordinates. The calculation process follows the orthogonality and linearity properties of the Fourier transform. After the transformation, the amplitude value of the frequency domain function is calculated; that is, the frequency domain amplitude spectrum of the structured light reflection image is obtained by taking the modulus of the complex form of the frequency domain function.
[0028] The energy peak coordinates of the frequency domain amplitude spectrum are retrieved using a peak detection algorithm. Calculate the frequency value of the energy peak coordinate point; In detail, a peak detection algorithm is used to traverse the frequency domain amplitude spectrum. A sliding window is used to compare the amplitude values of the center pixel with those of its neighboring pixels. When the amplitude value of the center pixel meets the condition of being greater than a preset multiple of the mean of the neighboring pixels and being a local maximum, it is determined to be an energy peak point, and its coordinates are recorded. Based on the sampling frequency of the frequency domain amplitude spectrum and the image resolution, and combined with the Fourier transform frequency mapping relationship, the coordinates are... This is converted into a corresponding frequency value, which reflects the dominant frequency characteristics of the structured light stripes.
[0029] by Centered on, radius Define the frequency domain region, extract the frequency components of the frequency domain region, and perform inverse Fourier transform on the frequency components to obtain the reconstructed fringe pattern. In detail, an inverse Fourier transform is performed on the extracted frequency components to convert the frequency domain signal back to the spatial domain, and a stripe pattern containing only the main frequency stripe information is reconstructed. This pattern can effectively filter out frequency domain noise and irrelevant frequency interference, highlighting the true shape of the structured light stripes.
[0030] The absolute phase value of each pixel in the reconstructed striped pattern is calculated using a spatial phase analysis algorithm. In detail, all pixels of the reconstructed stripe pattern are traversed. Based on the four-step phase-shifting algorithm and combined with the periodicity and frequency characteristics of the stripes, seed points for initial phase unfolding are determined. Pixel regions with stable phase gradients and low noise interference in the pattern are selected as the starting point. Starting from the seed point, based on the phase difference constraints between adjacent pixels, the theoretical phase difference range is calculated using the frequency information of the reconstructed stripes as an effective threshold for determining the phase difference between adjacent pixels. Phase unfolding is performed pixel by pixel. By comparing the phase relationship between the current pixel and the unfolded neighboring pixels, and combining the assumption of spatial domain continuity, the absolute phase value of the current pixel is calculated. During the unfolding process, phase unwrapping algorithms, such as the quality-guided path planning algorithm, are introduced to correct phase ambiguities caused by noise or stripe distortion, ensuring the continuity and accuracy of phase unfolding. Finally, the absolute phase distribution of each pixel in the reconstructed stripe pattern is obtained. This distribution can accurately reflect the stripe phase changes caused by the surface morphology of the ice aggregate, providing key phase data support for subsequent 3D contour extraction and dimensional parameter calculation of the ice aggregate.
[0031] Substituting the absolute phase value into the calibration formula, the PDR stripe frequency value of each pixel is calculated. The calibration formula is as follows:
[0032] In the formula, It is to reconstruct the pixels in the striped pattern. PDR fringe frequency values, In reconstructing the striped pattern, pixels The absolute phase value, In reconstructing the striped pattern, pixels The absolute phase value, It refers to the phase difference between adjacent pixels in the reconstructed stripe pattern. It refers to the physical spacing between adjacent pixels in the reconstructed stripe pattern; A PDR fringe frequency distribution map is generated based on the PDR fringe frequency values.
[0033] In detail, the PDR stripe frequency value is a key feature parameter of the pixels in the reconstructed stripe pattern. It is used to quantify the periodic variation characteristics of the stripes in the spatial domain. It is calculated based on the absolute phase difference and physical spacing of adjacent pixels, reflecting the number of stripe changes per unit physical length. It can accurately characterize the texture frequency distribution of the reconstructed stripe pattern. The PDR stripe frequency distribution map maps the PDR stripe frequency values of all pixels according to their spatial positions, and presents the distribution pattern of stripe frequencies in the detection area in a visual form. It provides global data support for the texture frequency distribution for subsequent analysis of ice aggregate characteristics and is a key medium for linking the stripe pattern with the physical characteristics of the detected object.
[0034] Specifically, a blank image canvas with the same pixel resolution as the reconstructed stripe pattern is constructed. The size and layout of its pixel array strictly match the reconstructed stripe pattern. Each pixel of the reconstructed stripe pattern is traversed, and the calculated PDR stripe frequency value is extracted. Following preset grayscale or color mapping rules (e.g., mapping low-frequency values to dark colors and high-frequency values to light colors), the mapping relationship is calibrated experimentally and stored in the algorithm parameter library. The frequency values are converted into the corresponding pixel's color or grayscale information and filled point-by-point onto the corresponding coordinate positions of the blank canvas. During the mapping process, if any frequency value exceeds a preset threshold range, the outlier correction module is invoked to perform interpolation or smoothing based on the frequency distribution of neighboring pixels, ensuring the continuity and accuracy of the distribution map. After mapping and correction of all pixels, a complete PDR stripe frequency distribution map is output, visually presenting the spatial distribution characteristics of the PDR stripe frequency within the detection area.
[0035] S3. Perform pixel-by-pixel weighted fusion calculation on the TCP signal feature image and the PDR stripe frequency distribution map to obtain the coupling strength distribution map; In an embodiment of the present invention, a pixel-by-pixel weighted fusion calculation is performed on the TCP signal feature image and the PDR stripe frequency distribution map to obtain a coupling strength distribution map, including: After spatially registering the TCP signal feature image and the PDR stripe frequency distribution map, extract the TCP cold light intensity value and PDR stripe frequency value of the pixels at the same position. Specifically, a registration algorithm based on feature point matching is used, such as extracting TCP cold light intensity feature points and PDR stripe frequency feature points from two images to construct feature point pairs. Then, the transformation matrix is solved using the least squares method to transform the TCP signal feature image and the PDR stripe frequency distribution map to the same spatial coordinate system, ensuring a one-to-one correspondence between their pixel positions. After spatial registration is completed, each pixel position in the registered image is traversed. Through an image pixel traversal mechanism, the TCP cold light intensity value in the TCP signal feature image at that position is extracted sequentially, i.e., the gray value of the corresponding pixel or the intensity parameter after feature quantization, and the PDR stripe frequency value in the PDR stripe frequency distribution map, i.e., the frequency parameter calculated and mapped and stored using the calibration formula mentioned above.
[0036] The coupling strength value is obtained by weighting the TCP cold light intensity value and the PDR fringe frequency value. The formula for calculating the coupling strength value is as follows:
[0037] In the formula, Pixels at the same position The coupling strength value at that point, Pixels at the same position TCP cold light intensity value at the location, Pixels at the same position PDR fringe frequency value at that location These are coupling weight coefficients; Generate a coupling strength distribution map based on the coupling strength value.
[0038] Specifically, after calculating the coupling strength values of all pixels, a blank canvas with the same resolution as the TCP signal feature image and the PDR stripe frequency distribution map is created. Each pixel position is traversed, and the coupling strength values are converted into color information and filled into the canvas according to preset mapping rules, such as low coupling values corresponding to cool colors and high values corresponding to warm colors. When abnormal values are encountered, interpolation correction is performed based on the normal distribution of the neighborhood. Finally, the coupling strength distribution map is output.
[0039] S4. Based on the statistical parameters of the coupling strength distribution map, the ice film area of the ice aggregate detection area is binarized to obtain the ice film mask; In an embodiment of the present invention, the ice film region of the ice aggregate detection area is binarized and determined according to the statistical parameters of the coupling intensity distribution map to obtain an ice film mask, including: The ice film detection threshold is calculated based on statistical parameters. The formula for calculating the ice film detection threshold is as follows:
[0040] In the formula, It is the threshold for determining the ice film. It is the average coupling strength in the statistical parameters. It is the standard deviation in the statistical parameters. It is the influence coefficient; In detail, first, all pixels in the coupling strength distribution map are traversed to obtain the coupling strength value of each pixel, forming a coupling strength dataset. The average coupling strength is obtained by summing all coupling strength values in the dataset and then dividing by the total number of pixels. The coupling strength value of each pixel and the mean are then calculated. The square of the difference is calculated by summing these squared values, dividing by the total number of pixels, and finally taking the square root of the result. .
[0041] If the coupling strength value at a pixel in the coupling strength distribution map is greater than the ice film determination threshold, the pixel is marked as an ice film region and assigned a value of 1; otherwise, the pixel is marked as a non-ice film region and assigned a value of 0. An ice mask is generated based on each pixel of the coupling strength distribution map.
[0042] In detail, the ice film area in the ice aggregate detection area refers to the area covered by ice film on or around the surface of the ice aggregate within the ice aggregate detection range. It is a specific object that needs to be identified and analyzed in ice aggregate detection. The ice film mask is an image generated by binary determination based on the statistical parameters of the coupling intensity distribution map. It accurately marks the distribution range of the ice film in the ice aggregate detection area in the form of pixel assignment.
[0043] Specifically, the coupling intensity distribution map is preprocessed, and a Gaussian filtering algorithm is used to eliminate random noise in the image while preserving edge information. Based on the preprocessed coupling intensity distribution map, a global threshold is calculated, and image pixels are divided into foreground and background categories. For each pixel, if its coupling intensity value is lower than the threshold, it is determined to be an ice film area, and the corresponding mask pixel value is set to 1; if it is higher than the threshold, it is determined to be a non-ice film area, and the corresponding mask pixel value is set to 0. After completing the judgment and assignment operations for all pixels in the coupling intensity distribution map, the assignment results of all pixels are integrated to generate an ice film mask that can accurately identify the distribution of ice film areas and non-ice film areas, providing clear and accurate mask data support for the subsequent analysis of ice aggregate detection areas.
[0044] S5. Generate aggregate candidate regions based on ultraviolet channel images and structured light reflection images, and calculate the actual size parameters of ice aggregates in the ice aggregate detection region based on the aggregate candidate regions and ice film mask. In an embodiment of the present invention, generating aggregate candidate regions based on ultraviolet channel images and structured light reflection images includes: Grayscale correction and Gaussian filtering noise reduction are performed on ultraviolet channel images and structured light reflection images; Specifically, grayscale correction is performed on the ultraviolet channel image and the structured light reflection image. By calculating the deviation between the mean grayscale value of the image and the mean grayscale value of the target image, a linear transformation model is used to adjust the grayscale values of the image pixels, so that the grayscale distribution of the two images tends to be consistent, thereby improving the compatibility between the images. Then, a Gaussian filtering algorithm is used to denoise the corrected image. Based on the image noise characteristics, an appropriate Gaussian kernel parameter is set so that the image can effectively filter out random noise while preserving edge information.
[0045] The pre-screening region of aggregates is extracted by edge detection, and candidate regions of aggregates are selected from the pre-screening region by connected component analysis.
[0046] Specifically, multi-scale Canny edge detection was performed on the preprocessed ultraviolet channel image and structured light reflectance image. A low threshold was set to 0.3 times the average grayscale value of the image, and a high threshold was set to 3 times the low threshold to generate an initial edge map. Morphological closure operations were performed on the edge map, first dilating and then eroding, and 5×5 structuring elements were used to fill the edge gaps to form a continuous closed contour, obtaining the aggregate pre-screening region. Eight-neighborhood connected component analysis was performed on the pre-screening region, marking all independent connected components and calculating their area, perimeter, circularity, and other geometric parameters. For each connected component, sequentially... Geometric feature screening is performed: areas with an area greater than 50 pixels and less than 10% of the total image area are retained, irregular areas with a circularity of less than 0.2 are excluded, and suspected aggregate areas with a rectangularity in the range of 0.5-0.9 are selected; for the still-adhesive areas after screening, the watershed algorithm combined with distance transformation is used for segmentation. By calculating the Euclidean distance transformation map of the area, the local maximum value of the area is determined as the seed point to achieve the separation of adhesive aggregates; the connected components that pass all screening conditions are determined as aggregate candidate areas, providing a basis for the accurate identification of ice aggregates in the future.
[0047] In embodiments of the present invention, the actual size parameters of the ice aggregate in the ice aggregate detection area are calculated based on the aggregate candidate area and the ice film mask, including: Morphological treatment of the ice film mask; Specifically, morphological processing is performed on the ice mask, and appropriate structural elements, such as 3×3 rectangular structural elements, are selected. Expansion and erosion operations are performed in sequence. The expansion operation can fill the tiny holes in the ice mask, making the outline of the ice mask area more complete, while the erosion operation is used to remove burrs and isolated noise points at the edge of the ice mask, thereby optimizing the mask quality.
[0048] Under the constraint of the aggregate candidate region, the contour of the morphologically processed ice film mask is extracted to obtain the real aggregate contour image; Specifically, the morphologically processed ice film mask image and aggregate candidate region image are loaded. Each pixel of the ice film mask image is traversed. Based on contour tracking algorithms, such as an eight-neighbor boundary traversal strategy, continuous edge pixels in the ice film mask are identified and marked to initially extract the contour information of the ice film region. The extracted ice film contour is then spatially intersected with the aggregate candidate region. By determining whether the contour pixels are within the aggregate candidate region, contour segments that are completely or partially within the aggregate candidate region are retained, while invalid contours outside the aggregate candidate region are filtered out. For cross-regional... The outline of the aggregate candidate region boundary is cropped according to the shape of the aggregate candidate region to ensure that the outline only includes the ice film portion within the aggregate candidate region. The cropped outline is then smoothed using polynomial fitting or spline curve interpolation to optimize the continuity and regularity of the outline. All the filtered, cropped, and smoothed outlines are integrated to generate a true aggregate outline image. This image accurately presents the true aggregate boundary shape corresponding to the ice film mask under the constraint of the aggregate candidate region, providing a reliable outline basis for the subsequent calculation of the actual size parameters of the ice aggregate.
[0049] The actual size parameters of the ice aggregate are calculated based on the actual aggregate outline image. The actual size parameters include the actual perimeter, actual length, actual width, actual area, actual aspect ratio, and actual shape factor.
[0050] Specifically, when calculating the actual size parameters of ice aggregate, the real aggregate outline image is first analyzed pixel by pixel, traversing all pixels on the outline, and the mapping relationship between the physical size and the pixel is determined in advance by a standard size calibrator, such as 1 pixel corresponding to the actual physical length. ), calculate the actual perimeter: accumulate the Euclidean distances between contour pixels, where the horizontal and vertical distances between each adjacent pixel are respectively and The diagonal distance is The actual perimeter is obtained by summing the distances of all adjacent points; the minimum bounding rectangle algorithm is used to traverse the contour pixels to determine the smallest rectangle that can enclose the contour. The length and width of the rectangle correspond to the actual length and width, and its length is multiplied by the number of pixels on the rectangle's side. The actual area is obtained by counting the number of pixels contained in the outline and then multiplying it by the actual area corresponding to a single pixel. Calculation; the actual aspect ratio is the ratio of the length to the width of the smallest bounding rectangle; the actual shape factor is based on the formula. The calculation comprehensively reflects the shape regularity of the aggregate.
[0051] S6. Calculate the freezing water correction factor for the average coupling intensity of the coupling intensity distribution map to obtain the freezing water content in the ice aggregate detection area; In detail, frozen water content refers to the proportion of water in the ice aggregate testing area that is not completely frozen. It is a key indicator reflecting the degree of freezing and distribution characteristics of water in the area and can be used to quantitatively assess the freezing and water content of ice aggregate.
[0052] In an embodiment of the present invention, the average coupling intensity of the coupling intensity distribution map is converted using a freezing water correction factor to obtain the freezing water content of the ice aggregate detection area, including: Calculate the average coupling strength of the coupling strength distribution map; In detail, all pixels in the coupling strength distribution map are traversed to obtain the coupling strength value of each pixel, forming a coupling strength dataset. The average coupling strength is obtained by summing all coupling strength values in the dataset and then dividing by the total number of pixels.
[0053] Substituting the average coupling strength into the corrected formula, the frozen water content is obtained. The calculation formula for the corrected formula is as follows:
[0054] In the formula, It is the frozen water content. It is the frozen water correction factor. It is the average coupling strength.
[0055] In detail, the freezing water content is calculated by substituting the average coupling strength into the correction formula. The principle is based on the inherent correlation between the distribution of freezing water and the coupling strength within the ice aggregate detection area. Extensive experiments and data analysis have verified that the presence of freezing water in the ice aggregate detection area alters the TCP cold light propagation and PDR stripe frequency distribution, causing a regular change in coupling strength. Furthermore, a stable linear correlation exists between the freezing water content and the average coupling strength. The freezing water correction coefficient is a key parameter determined through calibration experiments. Under standard freezing water content sample conditions, coupling strength distribution maps corresponding to different freezing water contents are collected. The average coupling strength is calculated and its relationship with the freezing water content is fitted to obtain a correction coefficient that accurately reflects the linear correlation between the two. Once the average coupling strength of the detection area is obtained, this correction formula can be used to quantify and convert the average coupling strength into freezing water content based on the established linear correspondence. This enables accurate calculation of the freezing water content within the ice aggregate detection area, providing a key quantitative indicator for ice aggregate characteristic analysis and related engineering applications.
[0056] S7. Determine whether the ice aggregate is qualified based on the frozen water content and actual size parameters.
[0057] In embodiments of the present invention, determining whether ice aggregate is qualified based on frozen water content and actual size parameters includes: The actual size parameters of all ice aggregates were grouped according to particle size range to obtain multiple particle size distribution groups; Specifically, the measurement index for ice aggregate particle size is clearly defined, and the actual length in the actual size parameters is selected as the grouping basis. Based on the application standards, industry specifications, or experimental requirements of ice aggregate, a reasonable particle size interval division scheme is determined. For example, it can be set according to continuous and non-overlapping intervals such as 0-5mm, 5-10mm, and 10-15mm. The specific values of the intervals can be determined by statistical analysis of the typical dimensions of the target ice aggregate. The actual size data of all ice aggregates are traversed, the actual length corresponding to each ice aggregate is extracted, and it is matched with the preset particle size interval to determine which particle size interval the size of the ice aggregate falls into. Then, it is divided into the corresponding group. After traversing and grouping all ice aggregates, multiple particle size distribution groups are obtained. Each group contains ice aggregates whose size parameters fall within the corresponding particle size interval.
[0058] The number of ice aggregates within each particle size distribution group and the total number of ice aggregates in the ice aggregate testing area are statistically analyzed. The ratio of the number of ice aggregates within each group to the total number of ice aggregates is calculated to obtain the proportion of ice aggregates in each particle size distribution group. Calculate the total area of all ice aggregates within the particle size distribution group based on the actual area, and calculate the ratio of the total area within the group to the total area of ice aggregates in the ice aggregate detection area to obtain the area ratio of ice aggregates in the particle size distribution group. For each particle size distribution group, the shape category of ice aggregate is determined based on the actual aspect ratio and actual shape factor. The proportion of ice aggregate quantity under each shape category is counted. The ratio of the proportion of ice aggregate quantity under each category to the total number of ice aggregates in the group is calculated to obtain the proportion of ice aggregate shape category in the particle size distribution group. The quality of ice aggregate is determined based on the proportion of aggregate quantity, ice aggregate area, ice aggregate shape category, and frozen water content.
[0059] Specifically, the proportion of aggregate quantity, ice aggregate area, and ice aggregate shape category for each particle size distribution group are compared one by one with the standard proportion range determined in advance through statistical analysis of a large number of qualified ice aggregate samples. This range covers the reasonable distribution range of qualified ice aggregate in terms of quantity, area, and shape category under different particle size intervals to determine whether each proportion indicator is within the standard range. The calculated frozen water content is compared with the standard threshold for frozen water content of qualified ice aggregate to determine whether the frozen water content meets the requirements. This threshold is determined experimentally based on the performance requirements and application scenarios of ice aggregate. Based on the comparison results of each indicator, if the proportion of aggregate quantity, ice aggregate area, and ice aggregate shape category for all particle size distribution groups are within the standard proportion range and the frozen water content does not exceed the standard threshold, the ice aggregate is judged to be qualified; otherwise, if at least one indicator exceeds the standard range or the frozen water content exceeds the standard, the ice aggregate is judged to be unqualified. In this way, the comprehensiveness and accuracy of ice aggregate quality assessment are ensured through multi-dimensional parameter collaborative judgment, meeting the strict requirements of actual application for ice aggregate quality.
[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for intelligent detection of aggregate gradation based on image recognition, characterized in that, The method includes: S1. Collect ultraviolet channel images and structured light reflection images of the ice aggregate detection area; S2. Extract signals from the ultraviolet channel image to obtain the TCP signal feature image, and perform stripe frequency analysis on the structured light reflection image to obtain the PDR stripe frequency distribution map. S3. Perform pixel-by-pixel weighted fusion calculation on the TCP signal feature image and the PDR stripe frequency distribution map to obtain the coupling strength distribution map; S4. Based on the statistical parameters of the coupling strength distribution map, the ice film area of the ice aggregate detection area is binarized to obtain the ice film mask; S5. Generate aggregate candidate regions based on ultraviolet channel images and structured light reflection images, and calculate the actual size parameters of ice aggregates in the ice aggregate detection region based on the aggregate candidate regions and ice film mask. S6. Calculate the freezing water correction factor for the average coupling intensity of the coupling intensity distribution map to obtain the freezing water content in the ice aggregate detection area; S7. Determine whether the ice aggregate is qualified based on the frozen water content and actual size parameters.
2. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, Acquire ultraviolet channel images and structured light reflectance images of the ice aggregate testing area, including: In response to the detection signal At any given moment, the TCP material in the ice aggregate detection area is excited by an ultraviolet pulse light source to generate cold light radiation, and an ultraviolet channel image of the ice aggregate detection area is acquired. exist At ms, striped structured light is projected onto the ice aggregate through a striped structured light projector, and the structured light reflection image of the ice aggregate detection area is acquired.
3. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, Signal extraction was performed on the ultraviolet channel image to obtain the TCP signal feature image. Stripe frequency analysis was performed on the structured light reflectance image to obtain the PDR stripe frequency distribution map, including: The background image of the area without aggregate is acquired, and pixel-by-pixel difference operation is performed on the ultraviolet channel image and the background image to obtain the TCP raw difference image; The TCP raw differential image is subjected to noise cancellation processing to obtain the TCP signal feature image; A two-dimensional Fourier transform is performed on the structured light reflection image to obtain the frequency domain amplitude spectrum of the structured light reflection image; The energy peak coordinates of the frequency domain amplitude spectrum are retrieved using a peak detection algorithm. Calculate the frequency value of the energy peak coordinate point; by Centered on, radius Define the frequency domain region, extract the frequency components of the frequency domain region, and perform inverse Fourier transform on the frequency components to obtain the reconstructed fringe pattern. The absolute phase value of each pixel in the reconstructed striped pattern is calculated using a spatial phase analysis algorithm. Substitute the absolute phase value into the calibration formula to calculate the PDR stripe frequency value for each pixel. A PDR fringe frequency distribution map is generated based on the PDR fringe frequency values.
4. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, A pixel-by-pixel weighted fusion calculation is performed on the TCP signal feature image and the PDR stripe frequency distribution map to obtain the coupling strength distribution map, including: After spatially registering the TCP signal feature image and the PDR stripe frequency distribution map, extract the TCP cold light intensity value and PDR stripe frequency value of the pixels at the same position. The coupling strength value is obtained by weighting the TCP cold light intensity value and the PDR fringe frequency value. Generate a coupling strength distribution map based on the coupling strength value.
5. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, Based on the statistical parameters of the coupling strength distribution map, the ice film area in the ice aggregate detection area is binarized to obtain the ice film mask, including: The threshold for determining ice film is calculated based on statistical parameters; If the coupling strength value at a pixel in the coupling strength distribution map is greater than the ice film determination threshold, the pixel is marked as an ice film region and assigned a value of 1; otherwise, the pixel is marked as a non-ice film region and assigned a value of 0. An ice mask is generated based on each pixel of the coupling strength distribution map.
6. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, Aggregate candidate regions are generated based on ultraviolet channel images and structured light reflectance images, including: Grayscale correction and Gaussian filtering noise reduction are performed on ultraviolet channel images and structured light reflection images; The pre-screening region of aggregates is extracted by edge detection, and candidate regions of aggregates are selected from the pre-screening region by connected component analysis.
7. The intelligent detection method for aggregate gradation based on image recognition as described in claim 1, characterized in that, The actual size parameters of ice aggregate in the ice aggregate detection area are calculated based on the aggregate candidate region and the ice film mask, including: Morphological treatment of the ice film mask; Under the constraint of the aggregate candidate region, the contour of the morphologically processed ice film mask is extracted to obtain the real aggregate contour image; The actual size parameters of the ice aggregate are calculated based on the actual aggregate outline image. The actual size parameters include the actual perimeter, actual length, actual width, actual area, actual aspect ratio, and actual shape factor.
8. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 1, characterized in that, The average coupling intensity of the coupling intensity distribution map is converted using a freezing water correction factor to obtain the freezing water content in the ice aggregate detection area, including: Calculate the average coupling strength of the coupling strength distribution map; Substituting the average coupling strength into the modified formula yields the frozen water content.
9. The intelligent detection method for aggregate gradation based on image recognition as described in claim 8, characterized in that, The corrected formula is calculated as follows: In the formula, It is the frozen water content. It is the frozen water correction factor. It is the average coupling strength.
10. The intelligent detection method for ice aggregate gradation based on image recognition as described in claim 7, characterized in that, Determining the quality of ice aggregate based on its frozen water content and actual size parameters includes: The actual size parameters of all ice aggregates were grouped according to particle size range to obtain multiple particle size distribution groups; The number of ice aggregates within each particle size distribution group and the total number of ice aggregates in the ice aggregate testing area are statistically analyzed. The ratio of the number of ice aggregates within each group to the total number of ice aggregates is calculated to obtain the proportion of ice aggregates in each particle size distribution group. Calculate the total area of all ice aggregates within the particle size distribution group based on the actual area, and calculate the ratio of the total area within the group to the total area of ice aggregates in the ice aggregate detection area to obtain the area ratio of ice aggregates in the particle size distribution group. For each particle size distribution group, the shape category of ice aggregate is determined based on the actual aspect ratio and actual shape factor. The proportion of ice aggregate quantity under each shape category is counted. The ratio of the proportion of ice aggregate quantity under each category to the total number of ice aggregates in the group is calculated to obtain the proportion of ice aggregate shape category in the particle size distribution group. The quality of ice aggregate is determined based on the proportion of aggregate quantity, ice aggregate area, ice aggregate shape category, and frozen water content.