A method and device for detecting warehouse wall cracks based on three-dimensional laser point cloud data

Abstract

This invention relates to the field of crack detection technology, and discloses a method and apparatus for detecting warehouse wall cracks based on three-dimensional laser point cloud data. The method includes: acquiring cross-sectional data of the target warehouse wall; performing spectral transformation on the cross-sectional data to obtain the cross-sectional data spectrum; acquiring three-dimensional laser point cloud data of the target warehouse wall; separating the warehouse wall control contour from the three-dimensional laser point cloud data based on the cross-sectional data spectrum; determining the texture distribution features of each cross-section of the target warehouse wall based on the control contour elevation; calculating a segmentation threshold for the cross-section based on the texture distribution features; determining suspected crack points for each cross-section based on the segmentation threshold and the control contour elevation; combining each suspected crack point to obtain a binary image representing the crack; performing region recognition on the binary image representing the crack to obtain crack regions; and detecting cracks in the target warehouse wall based on the crack regions. This invention can improve the accuracy of warehouse wall crack detection.

Description

Technical Field

[0001] This invention relates to the field of crack detection technology, and in particular to a method and apparatus for detecting warehouse wall cracks based on three-dimensional laser point cloud data. Background Technology

[0002] During the layered pouring of concrete in the silo, cracks may be caused by improper handling of the joints and the boundary between the old and new surfaces. If these cracks are not repaired in time, they may lead to safety accidents. In order to detect cracks in the silo wall in a timely manner, it is necessary to obtain information on the damage on the inside of the silo wall from time to time for timely inspection.

[0003] Existing crack detection technologies rely on deep learning to build crack detection models and then detect surface cracks. In practical applications, the data in silo wall crack detection varies slightly. Considering only surface cracks may lead to incomplete detection of the crack area, resulting in low accuracy in silo wall crack detection. Summary of the Invention

[0004] This invention provides a method and apparatus for detecting warehouse wall cracks based on three-dimensional laser point cloud data, the main purpose of which is to solve the problem of low accuracy in warehouse wall crack detection.

[0005] To achieve the above objectives, the present invention provides a method for detecting warehouse wall cracks based on three-dimensional laser point cloud data, comprising:

[0006] S1. Obtain the cross-sectional data of the target warehouse wall, and use the preset fast Fourier transform algorithm to perform spectral transformation on the cross-sectional data to obtain the cross-sectional data spectrum.

[0007] S2. Collect three-dimensional laser point cloud data of the target warehouse wall, and use a preset fast inverse Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data according to the cross-sectional data spectrum.

[0008] S3. Determine the texture distribution features of each cross section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile, and calculate the segmentation threshold of the cross section based on the texture distribution features, wherein calculating the segmentation threshold of the cross section based on the texture distribution features includes:

[0009] S31. Extract the elevation difference corresponding to the texture distribution features;

[0010] S32. Calculate the segmentation threshold of the cross section based on the elevation difference using the following segmentation threshold calculation formula:

[0011]

[0012] in, The segmentation threshold is... cross section The elevation difference of each measuring point This represents the total number of sampling points in a single cross-section. This is the threshold coefficient;

[0013] S4. Determine the suspected crack points of each cross section according to the segmentation threshold and the control contour elevation, and combine each suspected crack point to obtain a crack characterization binary image.

[0014] S5. Perform region identification on the binary image representing the crack to obtain the crack region, and use a preset minimum cost spanning tree algorithm to detect the cracks in the target warehouse wall based on the crack region.

[0015] Optionally, the step of performing a spectral transformation on the cross-sectional data using a preset fast Fourier transform algorithm to obtain the cross-sectional data spectrum includes:

[0016] Collect the cross-sectional analog signal corresponding to the cross-sectional data;

[0017] The cross-sectional analog signal is converted into a digital signal;

[0018] The digital signal is converted into a signal frequency amplitude spectrum using the Fast Fourier Transform algorithm, wherein the Fast Fourier Transform algorithm is as follows:

[0019]

[0020] in, Let k be the Fourier transform value of the k-th point in the digital signal. This represents the Fourier transform value of the k-th point in an even sequence of digital signals. This represents the Fourier transform value of the k-th point in an odd-numbered sequence of a digital signal. The length of the digital signal sequence. The rotation factor;

[0021] The signal power spectrum is determined based on the signal frequency amplitude spectrum, and the signal frequency amplitude spectrum and the signal power spectrum are used as the cross-sectional data spectrum.

[0022] Optionally, converting the cross-sectional analog signal into a digital signal includes:

[0023] The cross-sectional analog signal is used as the input value of a preset analog-to-digital converter, and it is determined whether the input value is within the standard value range of the analog-to-digital converter.

[0024] When the input value falls within the standard value range of the analog-to-digital converter, the conversion standard corresponding to the standard value range is obtained;

[0025] The cross-sectional analog signal is converted into a discrete signal represented by binary values ​​according to the conversion standard, and the discrete signal is used as the digital signal.

[0026] Optionally, after acquiring the three-dimensional laser point cloud data of the target warehouse wall, the method further includes:

[0027] A reference section corresponding to the cross-section of the target warehouse wall is determined using a preset median filtering algorithm;

[0028] Calculate the first distance between each breakpoint in the cross-section and the reference cross-section;

[0029] When the first distance is greater than or equal to a preset distance threshold, the breakpoint is regarded as an anomaly.

[0030] When the first distance is less than a preset distance threshold, the breakpoint is regarded as a non-abnormal point;

[0031] Calculate the second distance between each non-abnormal point and the abnormal point, and select the non-abnormal point with the smallest second distance to replace the abnormal point to obtain normal three-dimensional laser point cloud data.

[0032] Optionally, the step of using a preset inverse fast Fourier transform algorithm to separate the warehouse control contour in the three-dimensional laser point cloud data based on the cross-sectional data spectrum includes:

[0033] The inverse fast Fourier transform algorithm is used to calculate the functional relationship between the power spectrum and the preset transformed data in the cross-sectional data spectrum, wherein the frequency functional relationship is:

[0034]

[0035] in, The transformed data corresponding to the t-th point in the power spectrum. The length of the signal in the power spectrum. The sampling distance interval, Reference spatial frequency The power spectral density at that point For reference spatial frequency, For spatial frequency, It is the frequency index;

[0036] Extract the low-frequency band corresponding to the power spectrum based on the frequency function relationship;

[0037] The low-frequency band is truncated using a preset bandpass filter according to a preset low-frequency signal range value to obtain the bin wall control profile.

[0038] Optionally, determining the texture distribution features of each cross-section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile includes:

[0039] Obtain the cross-sectional profile elevation of the measuring points in each cross section;

[0040] The elevation difference between the cross-sectional profile elevation and the control profile elevation is calculated using the following formula:

[0041]

[0042] in, The first in the cross section The elevation difference of each measuring point The first in the cross section The cross-sectional profile elevation of each measuring point. The first in the cross section The control profile elevation of each measuring point;

[0043] The elevation difference is used as the texture distribution feature.

[0044] Optionally, determining the suspected crack points of each cross section based on the segmentation threshold and the control contour elevation includes:

[0045] Obtain the fluctuation value and crack elevation of each cross-sectional measuring point;

[0046] The measuring points whose fluctuation value is greater than the preset fluctuation threshold and whose crack elevation is less than the control contour elevation are designated as separate measuring points.

[0047] The crack point markers are determined using the following marking formula based on the elevation difference of the separated measuring points and the segmentation threshold:

[0048]

[0049] in, Mark the crack points. For the first The elevation difference of the separated measuring points The segmentation threshold is...

[0050] The suspected crack points of each cross section are determined based on the crack point markings.

[0051] Optionally, the step of performing region identification on the binary image representing the crack to obtain the crack region includes:

[0052] The binary image representing the crack is divided into sub-blocks to obtain sub-block images representing the crack.

[0053] The crack characterization sub-block image is subjected to confidence region filtering to obtain the crack confidence region;

[0054] The crack region is generated based on the crack confidence region.

[0055] Optionally, the step of detecting cracks in the target warehouse wall using a preset minimum cost spanning tree algorithm based on the crack region includes:

[0056] The crack region is refined to obtain crack seed points. All crack seed points are connected in pairs to obtain crack growth edges.

[0057] The marginal value of the pre-defined growth tree is calculated using the following formula:

[0058]

[0059] in, Seed point for crack With crack seed point The marginal value between them Seed point for crack With crack seed point The edges formed by vertices, For real numbers, Seed point for crack With crack seed point The length of the side between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The direction between, Seed point for crack With crack seed point The direction between, Pi Crack Seed Point With crack seed point The edges formed by vertices, Seed point for crack Match the seed point number of the crack at the other end corresponding to the solid edge;

[0060] The crack growth edge is allocated a cost value according to the edge cost value to obtain the crack growth cost edge, and the crack path is determined based on the crack growth cost edge using the minimum cost generation algorithm.

[0061] The crack path and the crack region are merged to obtain a merged crack region, and the cracks in the target warehouse wall are detected based on the merged crack region.

[0062] To address the aforementioned problems, the present invention also provides a warehouse wall crack detection device based on three-dimensional laser point cloud data, the device comprising:

[0063] The spectrum transformation module is used to acquire the cross-sectional data of the preset target warehouse wall, and to perform spectrum transformation on the cross-sectional data using a preset fast Fourier transform algorithm to obtain the spectrum of the cross-sectional data.

[0064] The warehouse wall control contour separation module is used to collect three-dimensional laser point cloud data of the target warehouse wall and use a preset fast inverse Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data based on the cross-sectional data spectrum.

[0065] The segmentation threshold calculation module is used to determine the texture distribution features of each cross section in the target warehouse wall based on the control contour elevation of the warehouse wall control contour, and to calculate the segmentation threshold of the cross section based on the texture distribution features.

[0066] The crack characterization binary image generation module is used to determine the suspected crack points of each cross section according to the segmentation threshold and the control contour elevation, and to combine each of the suspected crack points to obtain a crack characterization binary image.

[0067] The crack detection module is used to perform region identification on the binary image representing the crack to obtain the crack region, and to detect the cracks in the target warehouse wall based on the crack region using a preset minimum cost spanning tree algorithm.

[0068] This invention acquires three-dimensional laser point cloud data, uses frequency domain methods to separate the warehouse wall texture from the three-dimensional laser point cloud data, and realizes the warehouse wall control contour, thereby overcoming the influence of other damage to the warehouse wall texture and measurement posture, and improving the accuracy of warehouse wall crack detection. It then uses spatial domain methods to extract suspected crack data from the three-dimensional laser point cloud data, and combines each suspected crack data to generate a binary image representing the crack. This fully utilizes the continuity, direction, and clustering characteristics of the crack data, filtering out all data in the warehouse wall that may contain cracks, thus achieving complete and accurate detection of warehouse wall cracks. Therefore, the warehouse wall crack detection method and device based on three-dimensional laser point cloud data proposed in this invention can solve the problem of low accuracy in detecting warehouse wall cracks. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating a method for detecting warehouse wall cracks based on three-dimensional laser point cloud data, provided in an embodiment of the present invention.

[0070] Figure 2 A flowchart illustrating the control profile of the separation chamber wall according to an embodiment of the present invention;

[0071] Figure 3 This is a schematic diagram of a process for extracting texture distribution features according to an embodiment of the present invention;

[0072] Figure 4 This is a schematic diagram of a silo wall crack detection process provided in an embodiment of the present invention;

[0073] Figure 5 This is a functional block diagram of a warehouse wall crack detection device based on three-dimensional laser point cloud data provided in an embodiment of the present invention;

[0074] 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

[0075] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0076] This application provides a method for detecting warehouse wall cracks based on three-dimensional laser point cloud data. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for detecting warehouse wall cracks based on three-dimensional laser point cloud data can be executed by software or hardware installed on a terminal device or a server device. The software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0077] Reference Figure 1 The diagram shown is a flowchart illustrating a method for detecting warehouse wall cracks based on three-dimensional laser point cloud data according to an embodiment of the present invention. In this embodiment, the method for detecting warehouse wall cracks based on three-dimensional laser point cloud data includes S1-S5:

[0078] S1. Obtain the cross-sectional data of the target warehouse wall, and use the preset fast Fourier transform algorithm to perform spectral transformation on the cross-sectional data to obtain the cross-sectional data spectrum.

[0079] In this embodiment of the invention, the cross-section of the target warehouse wall refers to the section perpendicular to the centerline direction through the centerline stakes. The cross-section data refers to the measured ground elevation at the centerline stakes perpendicular to the centerline direction (normal direction). When performing cross-section measurements, the direction of the cross-section must first be determined. Since the ground elevation of each centerline stake has already been measured during the leveling measurement, it is only necessary to measure the distance and elevation difference between the ground change points on both sides of the centerline stakes and the stake points in this direction, thereby enabling the drawing of the cross-section diagram.

[0080] In detail, the direction of the cross section of the target warehouse wall can first be determined by using a compass or theodolite. The distance and elevation difference between the ground change points on both sides of the centerline stake and the stake point can be measured by using a level and tape measure, thereby determining the elevation of the cross section.

[0081] Furthermore, in the spatiotemporal domain of the signal, the warehouse wall texture, cracks, and control contours of the target warehouse wall overlap and influence each other, making it difficult to accurately extract and locate the control contours and crack information. Therefore, the cross-sectional data is subjected to spectral transformation to achieve accurate extraction of the warehouse wall control contours.

[0082] In this embodiment of the invention, a cross-sectional map is generated based on the horizontal distance and elevation of the cross-section data of the warehouse wall. Radiation source signals are collected based on the horizontal distance of the cross-section, and then the frequency and amplitude information of the radiation source signals are continuously obtained over a period of time using a Fast Fourier Transform (FFT) algorithm. The essence of the Fast Fourier Transform (FFT) algorithm lies in appropriately decomposing the calculation of the Discrete Fourier Transform (DFT) of a long sequence into the calculation of the DFT of a short sequence.

[0083] In this embodiment of the invention, the step of using a preset Fast Fourier Transform (FFT) algorithm to perform spectral transformation on the cross-sectional data to obtain the cross-sectional data spectrum includes: acquiring the cross-sectional analog signal corresponding to the cross-sectional data and converting the cross-sectional analog signal into a digital signal; wherein, the cross-sectional analog signal can be acquired based on the relationship between horizontal distance and elevation in the two-dimensional relationship diagram of the cross-sectional horizontal distance and cross-sectional elevation in the cross-sectional data, and since the cross-sectional analog signal is continuous, it is not suitable to use the Fourier Transform algorithm to convert its signal frequency amplitude spectrum and power spectrum. Therefore, it is necessary to convert the cross-sectional analog signal into a digital signal; the digital signal is converted into a signal frequency amplitude spectrum using the following Fast Fourier Transform algorithm, wherein the Fast Fourier Transform algorithm is:

[0084]

[0085] in, Let k be the Fourier transform value of the k-th point in the digital signal. This represents the Fourier transform value of the k-th point in an even sequence of digital signals. This represents the Fourier transform value of the k-th point in an odd-numbered sequence of a digital signal. The length of the digital signal sequence. The rotation factor;

[0086] The signal power spectrum is determined based on the signal frequency amplitude spectrum, and the signal frequency amplitude spectrum and the signal power spectrum are used as the cross-sectional data spectrum.

[0087] Specifically, the step of converting the cross-sectional analog signal into a digital signal includes: using the cross-sectional analog signal as the input value of a preset analog-to-digital converter, and determining whether the input value falls within the standard value range of the analog-to-digital converter; wherein, the analog-to-digital converter refers to a type of device used to convert a continuous signal in analog form into a discrete signal in digital form, and the digital signal refers to a signal in which both the independent and dependent variables are discrete; when the input value falls within the standard value range of the analog-to-digital converter, obtaining the conversion standard corresponding to the standard value range; converting the cross-sectional analog signal into a discrete signal represented by binary values ​​according to the conversion standard, and using the discrete signal as the digital signal.

[0088] For example, determining whether the input value belongs to the standard value range of the analog-to-digital converter means that when the input value is 15, the standard value range is 1 to 50, and 15 exists within 1 to 50; when the input value belongs to the standard value range of the analog-to-digital converter, obtaining the conversion standard corresponding to the standard value range means that when the standard value range is 1 to 50, the conversion standard corresponding to the input value 15 can be the conversion of the cross-sectional analog signal 15 to the digital signal 150.

[0089] Specifically, after being converted into a digital signal, the signal expression that varies with time is transformed into a frequency-varying spectral function expression using the Fast Fourier Transform (FFT) algorithm. Spectral analysis then identifies the frequency components within the signal, determining its amplitude and phase. The Fast Fourier Transform algorithm... It plays a rotating role in complex number multiplication, and is therefore called the rotation factor.

[0090] Furthermore, the Fast Fourier Transform (FFT) algorithm can transform the cross-sectional data of the warehouse wall into the signal frequency amplitude spectrum and power spectrum. In addition, noise is generated during signal transmission, and noise lacks corresponding spectral information. Moreover, the signal received at the receiver is never exactly the same as the signal transmitted; it is a random signal. The Fourier Transform, however, is designed for deterministic signals. That is, the spectrum of a random signal does not exist, but most random signals are stationary, and the power spectrum of a stationary random signal does exist. Therefore, it is necessary to determine the signal power spectrum.

[0091] Furthermore, within the signal frequency domain, since the warehouse wall texture represents high-frequency changing information, and the cracks also contain relatively abrupt changes, while the change trend of the warehouse wall control profile is relatively gentle. The cracks in the warehouse wall texture domain correspond to the high-frequency part of the spectrum, while the warehouse wall control profile corresponds to the low-frequency part. Therefore, using Fast Fourier Transform (FFT) can effectively separate the warehouse wall texture and the warehouse wall control profile, thereby enabling the extraction of the frequency band of the warehouse wall control profile.

[0092] S2. Collect three-dimensional laser point cloud data of the target warehouse wall, and use a preset inverse fast Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data according to the cross-sectional data spectrum.

[0093] In this embodiment of the invention, the three-dimensional laser point cloud data includes information such as the control contour of the warehouse wall, damage texture, attitude, and abnormal data. Among them, the damage and texture of the warehouse wall control contour are the basic components of the warehouse wall under ideal measurement conditions, while the attitude abnormal data are noise introduced by the actual measurement environment, including dynamic measurement environment, abnormal reflection of warehouse wall material, abnormal sensor data, measurement system errors, etc.

[0094] In detail, the three-dimensional laser point cloud data of the target warehouse wall can be collected by a laser 3D scanner. The laser 3D scanner uses the principle of laser ranging to record the three-dimensional coordinates, reflectivity and texture information of a large number of dense points on the surface of the object being measured, and can quickly reconstruct the three-dimensional model of the target and various graphic data such as lines, surfaces and volumes.

[0095] In this embodiment of the invention, when measuring 3D laser point cloud data of a warehouse wall in a moving environment, the obtained 3D laser point cloud data contains measurement attitude information. Usually, the attitude of each cross section is different. In global data processing, attitude has a significant impact on data consistency analysis, especially when data analysis is required along the measurement direction. Therefore, by simultaneously collecting the measurement attitude during data measurement, the collected measurement attitude data can be used to correct the attitude of the point cloud data.

[0096] Furthermore, due to uneven silo wall paving materials, abnormal reflections caused by water and oil stains, and the influence of foreign objects, the acquired 3D laser cross-sectional signals contain abnormal data caused by the above factors. Abnormal data in the cross-section typically exhibits strong randomness, usually appearing as large-amplitude pulse signals in point cloud data. The presence of abnormal data affects the accuracy of crack detection; therefore, to improve crack detection accuracy, it is necessary to reduce or eliminate these abnormal data through relevant preprocessing before data processing.

[0097] In this embodiment of the invention, after acquiring the three-dimensional laser point cloud data of the target warehouse wall, the method further includes: determining a reference cross-section corresponding to the cross-section of the target warehouse wall using a preset median filtering algorithm. The median filtering is a nonlinear signal processing method, a nonlinear filter, and also a statistical sorting filter. The grayscale value of each pixel is set to the median of the grayscale values ​​of all pixels within a certain neighborhood window of that pixel. The median filtering is used to select the noise-smoothed cross-section corresponding to the cross-section of the target warehouse wall as the reference cross-section. A first distance is calculated between each breakpoint in the cross-section and the reference cross-section. When the first distance is greater than or equal to a preset distance threshold, the breakpoint is considered an anomaly. When the first distance is less than the preset distance threshold, the breakpoint is considered a non-anomaly. A second distance is calculated between each non-anomaly and the anomaly. The non-anomaly with the smallest second distance is selected to replace the anomaly, thus obtaining normal three-dimensional laser point cloud data.

[0098] Specifically, the distance between each breakpoint in the cross-section of the target warehouse wall and the reference cross-section is calculated. Anomalies that deviate significantly from the reference cross-section are identified, and non-anomaly data near these anomalies are used to replace them. In other words, the abnormal data in the three-dimensional laser point cloud data corresponding to the target warehouse wall is replaced with non-anomaly data to improve the detection accuracy of cracks.

[0099] In this embodiment of the invention, in the point cloud data spectrum after inverse fast Fourier transform (FFT) (assuming the maximum value of the spectrum range is fH), the warehouse wall control profile is a smooth profile that fits the warehouse wall, corresponding to the low-frequency part of the spectrum. A bandpass filter can be used to truncate the frequency band of the warehouse wall control profile. The inverse fast Fourier transform (IFFT) process is the opposite of the frequency domain signal to the spatiotemporal domain conversion. The corresponding signal waveform is obtained by processing the power spectrum of the corresponding low-frequency band using IFFT, which is the warehouse wall control profile. The truncated low-frequency signal range is 0~fs1.

[0100] In this embodiment of the invention, the reference Figure 2As shown, the step of using a preset inverse fast Fourier transform algorithm to separate the warehouse control contour from the three-dimensional laser point cloud data based on the cross-sectional data spectrum includes:

[0101] S21. Calculate the functional relationship between the power spectrum and preset transformed data in the cross-sectional data spectrum using the inverse fast Fourier transform algorithm, wherein the frequency functional relationship is:

[0102]

[0103] in, The transformed data corresponding to the t-th point in the power spectrum. The length of the signal in the power spectrum. The sampling distance interval, Reference spatial frequency The power spectral density at that point For reference spatial frequency, For spatial frequency, It is the frequency index;

[0104] In detail, by using the inverse fast Fourier transform to convert the frequency domain function to the time domain function or the spatial domain function, and by converting the power spectrum in the cross-sectional data spectrum to the time domain or the spatial domain, the relationship between the power spectrum and the time domain or the spatial domain can be determined. That is, the inverse fast Fourier transform is used to determine its functional relationship.

[0105] S22. Extract the low-frequency band corresponding to the power spectrum according to the frequency function relationship;

[0106] S23. Using a preset bandpass filter, the signal waveform of the low-frequency band is truncated according to the preset low-frequency signal range value to obtain the bin wall control profile.

[0107] Specifically, the low-frequency band corresponding to the power spectrum is determined using the aforementioned functional relationship, and the signal waveform corresponding to the low-frequency band is truncated using a bandpass filter. The bandpass filter is a device that allows waves of a specific frequency band to pass through while blocking other frequency bands. The signal waveform of the low-frequency band is truncated using a preset low-frequency signal range value to obtain the silo control profile.

[0108] Furthermore, the low-frequency signal ranged from 0 to fs1. When the cutoff frequency fs1 = 0.003fH, the signal after bandpass filtering contained fewer components, meaning the recovered control profile was smoother, with some deviations from the cross-sectional profile trend. The recovered control profile at fs1 = 0.015fH closely followed the silo wall trend, with no loss of crack information. The recovered control profile at fs1 = 0.03fH closely followed the silo wall trend, but some crack information was lost. To facilitate accurate crack extraction, fs1 = 0.015fH was selected to obtain the silo wall control profile.

[0109] Furthermore, the actual texture value of the silo wall paving differs from the designed texture value, and impacts and friction during silo wall use cause changes in the silo wall texture. Therefore, it is necessary to analyze the texture distribution characteristics of each cross section to improve the accuracy of crack detection.

[0110] S3. Determine the texture distribution features of each cross section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile, and calculate the segmentation threshold of the cross section based on the texture distribution features.

[0111] In this embodiment of the invention, the silo wall contour data includes three parts: silo wall control contour, silo wall texture, and cracks. The elevation of the crack data is lower than that of the control contour. Ideally, the silo walls all have known design texture values. After separating the silo wall control contour from the silo wall contour, the silo wall contour data still contains silo wall texture and crack information. Combining the known design texture, the silo wall crack data can be obtained directly. However, the actual texture value of the silo wall paving differs from the design texture value, and impacts and friction during silo wall use cause changes in the silo wall texture. The silo wall texture directly reflects the local fluctuation amplitude of the point cloud data; the larger the texture, the greater the fluctuation, and the higher the segmentation threshold required to separate crack points in the data. Therefore, it is necessary to analyze the texture distribution characteristics of each cross section to obtain the silo wall crack data.

[0112] In this embodiment of the invention, the reference Figure 3 As shown, determining the texture distribution features of each cross-section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile includes:

[0113] S31. Obtain the cross-sectional profile elevation of the measuring points in each cross section, wherein the cross-sectional profile elevation refers to the pre-processed cross-sectional profile elevation, which can be obtained by a laser 3D scanner.

[0114] S32. Calculate the elevation difference between the cross-sectional profile elevation and the control profile elevation using the following elevation difference calculation formula:

[0115]

[0116] in, The first in the cross section The elevation difference of each measuring point The first in the cross section The cross-sectional profile elevation of each measuring point. The first in the cross section The control profile elevation of each measuring point;

[0117] Specifically, the local fluctuations in the silo wall profile are mainly caused by changes in the silo wall texture elevation. The elevation difference between the profile elevation of the measuring point in the preprocessed cross section and the control profile elevation can reflect the texture distribution characteristics of the cross section. Then, based on the texture distribution characteristics, the dynamic segmentation threshold of each cross section is determined, thereby realizing the extraction of suspected crack points of all cross sections and improving the accuracy and comprehensiveness of crack point extraction.

[0118] S33. The elevation difference is used as the texture distribution feature.

[0119] In this embodiment of the invention, calculating the segmentation threshold of the cross-section based on the texture distribution features includes:

[0120] Extract the elevation difference corresponding to the texture distribution features;

[0121] The segmentation threshold of the cross section is calculated based on the elevation difference using the following segmentation threshold calculation formula:

[0122]

[0123] in, The segmentation threshold is... cross section The elevation difference of each measuring point This represents the total number of sampling points in a single cross-section. This is the threshold coefficient.

[0124] In detail, statistical analysis of the elevation differences corresponding to the texture distribution features yields the texture mean and texture root mean square error. The first half of the segmentation threshold calculation formula... This represents the texture mean, the latter half. This represents the mean squared error of the texture. In addition, the threshold coefficient is usually between 2 and 3. Therefore, the segmentation threshold of the cross section can be calculated based on the mean texture and the mean squared error of the texture.

[0125] Furthermore, suspected crack points on the cross section can be segmented according to the segmentation threshold, thereby determining suspected crack points on all cross sections and ensuring the comprehensiveness of crack detection.

[0126] S4. Determine the suspected crack points of each cross section based on the segmentation threshold and the control contour elevation, and combine each suspected crack point to obtain a binary image representing the crack.

[0127] In this embodiment of the invention, since the fluctuation amplitude of crack points is greater than the average texture value, and the elevation of crack data is less than the elevation of control contour, suspected crack data is calculated based on the segmentation threshold and the elevation of the control contour of the silo wall, thereby determining the suspected crack points of all sections and ensuring the accuracy of crack detection.

[0128] In this embodiment of the invention, determining the suspected crack points of each cross section based on the segmentation threshold and the control contour elevation includes: acquiring the fluctuation value and crack elevation of each cross section measuring point; identifying measuring points whose fluctuation value is greater than a preset fluctuation threshold and whose crack elevation is less than the control contour elevation as separated measuring points; and determining crack point markings based on the elevation difference of the separated measuring points and the segmentation threshold using the following marking formula:

[0129]

[0130] in, Mark the crack points. For the first The elevation difference of the separated measuring points The segmentation threshold is...

[0131] The suspected crack points of each cross section are determined based on the crack point markings.

[0132] In detail, firstly, points with large fluctuation values ​​and values ​​smaller than the control contour in each cross-section measurement point are screened out, and these points are used as the separation measurement points. Then, the separation measurement points are separated into suspected crack points by the segmentation threshold T.

[0133] Specifically, if the elevation difference of the separated measuring points is greater than the segmentation threshold, the separated measuring point is marked as 1 and determined to be a crack point; if the elevation difference of the separated measuring points is less than or equal to the segmentation threshold, the separated measuring point is marked as 0 and determined to be a non-crack point. Suspected crack points on all cross-sections are then determined according to the marking rules in the marking formula.

[0134] In this embodiment of the invention, each suspected crack point is combined to obtain a crack characterization binary image. That is, the crack characterization binary image is generated by treating each suspected crack point as a pixel and collecting all pixels to form a binary image. Here, a binary image means that each pixel in the image has only two possible values ​​or grayscale levels.

[0135] Furthermore, based on the analysis of binary images representing cracks, crack identification is achieved to improve the completeness of crack detection.

[0136] S5. Perform region identification on the binary image representing the crack to obtain the crack region, and use a preset minimum cost spanning tree algorithm to detect the cracks in the target warehouse wall based on the crack region.

[0137] In this embodiment of the invention, the binary image obtained based on suspected crack data contains most of the crack data and some noise data. The crack data has certain clustering characteristics, which are manifested as having certain geometric shapes, including length, width and area. Therefore, it is necessary to analyze the geometric shape of the binary image representing the crack in order to achieve the completeness of crack detection.

[0138] In this embodiment of the invention, the step of performing region identification on the binary image representing the crack to obtain a crack region includes: dividing the binary image representing the crack into sub-blocks to obtain sub-block images representing cracks; and performing confidence region filtering on the sub-block images representing cracks to obtain crack confidence regions. Dividing the binary image representing the crack into sub-blocks, then evaluating the sub-blocks, filtering confidence regions, and dividing the image into sub-blocks can improve the accuracy of confidence region filtering. A preset random walk algorithm can be used to calculate the confidence level, and then the confidence region of the sub-block images representing cracks can be filtered based on the confidence level to obtain crack confidence regions; the crack region is then generated based on the crack confidence regions.

[0139] Specifically, suspected crack points with relatively long connected regions are extracted from the crack confidence region as the original crack seed region, and a minimum cost spanning tree crack growth method is adopted based on the crack seed region to further achieve complete crack detection.

[0140] In this embodiment of the invention, the step of detecting cracks in the target warehouse wall using a preset minimum cost spanning tree algorithm based on the crack region includes: refining the crack region to obtain crack seed points; connecting all the crack seed points pairwise to obtain crack growth edges; and calculating the edge cost value of the preset growth tree using the following cost value calculation formula:

[0141]

[0142] in, Seed point for crack With crack seed point The marginal value between them Seed point for crack With crack seed point The edges formed by vertices, For real numbers, Seed point for crack With crack seed point The length of the side between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The direction between, Seed point for crack With crack seed point The direction between, Pi Crack Seed Point With crack seed point The edges formed by vertices, Seed point for crack Match the seed point number of the crack at the other end corresponding to the solid edge; allocate the cost value of the crack growth edge according to the edge cost value to obtain the crack growth cost edge; determine the crack path according to the crack growth cost edge using the minimum cost generation algorithm; merge the crack path with the crack region to obtain the merged crack region; and detect the cracks in the target warehouse wall according to the merged crack region.

[0143] In detail, the original crack seed region is refined to obtain the skeleton of the original crack seed region; the directional turning points of the original crack seed region skeleton are extracted using the corner detection method, and the original crack seed region skeleton is adaptively segmented; the segmented crack segments are used as crack seed regions, and the endpoints of the crack seed regions are used as crack seed points.

[0144] Specifically, all crack seed points are connected pairwise. If two crack seed points belong to the same crack seed region, the connection is replaced by the crack seed region itself and marked as a solid edge; otherwise, it is marked as a dashed edge. The set of solid edges is denoted by R, and the set of dashed edges is denoted by V. Then, edge values ​​are assigned. Higher confidence results in lower edge values, and lower confidence results in higher edge values. Finally, crack paths are obtained based on the minimum cost spanning tree principle. These crack paths are then merged with the crack seed regions to achieve complete crack growth.

[0145] Furthermore, the cracks in the target warehouse wall are detected according to the fusion crack region, taking into account the characteristics of crack continuity, directionality, and aggregation, thereby achieving complete and accurate detection of the warehouse wall.

[0146] Furthermore, such as Figure 4The diagram illustrates the process of detecting cracks in the warehouse wall. Image (a) shows the depth map after preprocessing the cracks in the warehouse wall. The depth map is obtained by preprocessing the outliers in the original crack data. Based on the preprocessed depth map, suspected crack data is determined. A binary crack map is generated based on the suspected data, and image (b) shows the suspected crack binary map. The suspected crack binary map is then divided into sub-blocks to obtain the crack sub-blocks of the warehouse wall crack data, as shown in image (c). The classified crack sub-blocks are then extended by morphological sub-block sets to ensure the integrity of the warehouse wall crack detection, as shown in image (d). Finally, the cracks are grown and restored based on the minimum spanning tree to obtain a complete and accurate crack, as shown in image (e). This completes the detection of cracks in the warehouse wall, ensuring both the integrity and accuracy of the detection.

[0147] This invention acquires three-dimensional laser point cloud data, uses frequency domain methods to separate the warehouse wall texture from the three-dimensional laser point cloud data, and realizes the warehouse wall control contour, thereby overcoming the influence of other damage to the warehouse wall texture and measurement posture, and improving the accuracy of warehouse wall crack detection. It then uses spatial domain methods to extract suspected crack data from the three-dimensional laser point cloud data, and combines each suspected crack data to generate a binary image representing the crack. This fully utilizes the continuity, direction, and clustering characteristics of the crack data, filtering out all data in the warehouse wall that may contain cracks, thus achieving complete and accurate detection of warehouse wall cracks. Therefore, the warehouse wall crack detection method and device based on three-dimensional laser point cloud data proposed in this invention can solve the problem of low accuracy in detecting warehouse wall cracks.

[0148] like Figure 5 The diagram shown is a functional block diagram of a warehouse wall crack detection device based on three-dimensional laser point cloud data provided in an embodiment of the present invention.

[0149] The warehouse wall crack detection device 100 based on three-dimensional laser point cloud data described in this invention can be installed in an electronic device. Depending on the functions implemented, the warehouse wall crack detection device 100 based on three-dimensional laser point cloud data may include a spectrum transformation module 101, a warehouse wall control contour separation module 102, a segmentation threshold calculation module 103, a crack characterization binary image generation module 104, and a crack detection module 105. The module described in this invention can also be called a unit, referring to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0150] In this embodiment, the functions of each module / unit are as follows:

[0151] The spectrum transformation module 101 is used to acquire the preset cross-sectional data of the target warehouse wall, and to perform spectrum transformation on the cross-sectional data using a preset fast Fourier transform algorithm to obtain the cross-sectional data spectrum.

[0152] The warehouse wall control contour separation module 102 is used to collect three-dimensional laser point cloud data of the target warehouse wall and use a preset fast inverse Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data according to the cross-sectional data spectrum.

[0153] The segmentation threshold calculation module 103 is used to determine the texture distribution features of each cross section in the target warehouse wall based on the control contour elevation of the warehouse wall control contour, and to calculate the segmentation threshold of the cross section based on the texture distribution features.

[0154] The crack characterization binary image generation module 104 is used to determine the suspected crack points of each cross section according to the segmentation threshold and the control contour elevation, and combine each of the suspected crack points to obtain a crack characterization binary image.

[0155] The crack detection module 105 is used to perform region recognition on the binary image representing the crack to obtain the crack region, and to detect the cracks in the target warehouse wall based on the crack region using a preset minimum cost spanning tree algorithm.

[0156] In detail, the modules in the warehouse wall crack detection device 100 based on three-dimensional laser point cloud data described in this embodiment of the invention employ the same methods as described above during use. Figures 1 to 3 The method described herein uses the same techniques as the warehouse wall crack detection method based on three-dimensional laser point cloud data and can produce the same technical effect, so it will not be repeated here.

[0157] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0158] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0159] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0160] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0161] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0162] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0163] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices described in the system embodiments may also be implemented by a single unit or device through software or hardware. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0164] 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 detecting warehouse wall cracks based on three-dimensional laser point cloud data, characterized in that, The method includes: S1. Obtain the cross-sectional data of the target warehouse wall, and use the preset fast Fourier transform algorithm to perform spectral transformation on the cross-sectional data to obtain the cross-sectional data spectrum. S2. Collect three-dimensional laser point cloud data of the target warehouse wall, and use a preset fast inverse Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data according to the cross-sectional data spectrum. S3. Determine the texture distribution features of each cross section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile, and calculate the segmentation threshold of the cross section based on the texture distribution features, wherein calculating the segmentation threshold of the cross section based on the texture distribution features includes: S31. Extract the elevation difference corresponding to the texture distribution features; S32. Calculate the segmentation threshold of the cross section based on the elevation difference using the following segmentation threshold calculation formula: in, The segmentation threshold is... The first in the cross section The elevation difference of each measuring point This represents the total number of sampling points in a single cross-section. This is the threshold coefficient; S4. Determine the suspected crack points of each cross section according to the segmentation threshold and the control contour elevation, and combine each suspected crack point to obtain a crack characterization binary image. S5. Perform region identification on the binary image representing the crack to obtain the crack region, and use a preset minimum cost spanning tree algorithm to detect the cracks in the target warehouse wall based on the crack region.

2. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The step of performing a spectral transformation on the cross-sectional data using a preset Fast Fourier Transform algorithm to obtain the cross-sectional data spectrum includes: Collect the cross-sectional analog signal corresponding to the cross-sectional data; The cross-sectional analog signal is converted into a digital signal; The digital signal is converted into a signal frequency amplitude spectrum using the Fast Fourier Transform algorithm, wherein the Fast Fourier Transform algorithm is as follows: in, Let k be the Fourier transform value of the k-th point in the digital signal. This represents the Fourier transform value of the k-th point in an even sequence of digital signals. This represents the Fourier transform value of the k-th point in an odd-numbered sequence of a digital signal. The length of the digital signal sequence. The rotation factor; The signal power spectrum is determined based on the signal frequency amplitude spectrum, and the signal frequency amplitude spectrum and the signal power spectrum are used as the cross-sectional data spectrum.

3. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 2, characterized in that, The process of converting the cross-sectional analog signal into a digital signal includes: The cross-sectional analog signal is used as the input value of a preset analog-to-digital converter, and it is determined whether the input value is within the standard value range of the analog-to-digital converter. When the input value falls within the standard value range of the analog-to-digital converter, the conversion standard corresponding to the standard value range is obtained; The cross-sectional analog signal is converted into a discrete signal represented by binary values ​​according to the conversion standard, and the discrete signal is used as the digital signal.

4. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, After acquiring the three-dimensional laser point cloud data of the target warehouse wall, the method further includes: A reference section corresponding to the cross-section of the target warehouse wall is determined using a preset median filtering algorithm; Calculate the first distance between each breakpoint in the cross-section and the reference cross-section; When the first distance is greater than or equal to a preset distance threshold, the breakpoint is regarded as an anomaly. When the first distance is less than a preset distance threshold, the breakpoint is regarded as a non-abnormal point; Calculate the second distance between each non-abnormal point and the abnormal point, and select the non-abnormal point with the smallest second distance to replace the abnormal point to obtain normal three-dimensional laser point cloud data.

5. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The step of using a preset inverse fast Fourier transform algorithm to separate the warehouse control contour from the three-dimensional laser point cloud data based on the cross-sectional data spectrum includes: The inverse fast Fourier transform algorithm is used to calculate the functional relationship between the power spectrum and the preset transformed data in the cross-sectional data spectrum, wherein the frequency functional relationship is: in, The transformed data corresponding to the t-th point in the power spectrum. The length of the signal in the power spectrum. The sampling distance interval, Reference spatial frequency The power spectral density at that point For reference spatial frequency, For spatial frequency, It is the frequency index; Extract the low-frequency band corresponding to the power spectrum based on the frequency function relationship; The signal waveform of the low-frequency band is truncated using a preset bandpass filter according to a preset low-frequency signal range value to obtain the bin wall control profile.

6. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in any one of claims 1 to 5, characterized in that, The step of determining the texture distribution features of each cross section in the target warehouse wall based on the control profile elevation of the warehouse wall control profile includes: Obtain the cross-sectional profile elevation of the measuring points in each cross section; The elevation difference between the cross-sectional profile elevation and the control profile elevation is calculated using the following formula: in, The first in the cross section The elevation difference of each measuring point The first in the cross section The cross-sectional profile elevation of each measuring point. The first in the cross section The control profile elevation of each measuring point; The elevation difference is used as the texture distribution feature.

7. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The step of determining the suspected crack points of each cross section based on the segmentation threshold and the control contour elevation includes: Obtain the fluctuation value and crack elevation of each cross-sectional measuring point; The measuring points whose fluctuation value is greater than the preset fluctuation threshold and whose crack elevation is less than the control contour elevation are designated as separate measuring points. The crack point markers are determined using the following marking formula based on the elevation difference of the separated measuring points and the segmentation threshold: in, Mark the crack points. For the first The elevation difference of the separated measuring points The segmentation threshold is... The suspected crack points of each cross section are determined based on the crack point markings.

8. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The step of performing region identification on the binary image representing the crack to obtain the crack region includes: The binary image representing the crack is divided into sub-blocks to obtain sub-block images representing the crack. The crack characterization sub-block image is subjected to confidence region filtering to obtain the crack confidence region; The crack region is generated based on the crack confidence region.

9. The method for detecting warehouse wall cracks based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The step of detecting cracks in the target warehouse wall using a preset minimum cost spanning tree algorithm based on the crack region includes: The crack region is refined to obtain crack seed points. All crack seed points are then connected in pairs to obtain crack growth edges. The marginal value of the pre-defined growth tree is calculated using the following formula: in, Seed point for crack With crack seed point The marginal value between them Seed point for crack With crack seed point The edges formed by vertices, For real numbers, Seed point for crack With crack seed point The length of the side between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The normalization coefficients between them Seed point for crack With crack seed point The direction between, Seed point for crack With crack seed point The direction between, Pi Crack Seed Point With crack seed point The edges formed by vertices, Seed point for crack Match the seed point number of the crack at the other end corresponding to the solid edge; The crack growth edge is allocated a cost edge according to the edge cost value to obtain the crack growth cost edge, and the crack path is determined based on the crack growth cost edge using the minimum cost generation algorithm. The crack path and the crack region are merged to obtain a merged crack region, and the cracks in the target warehouse wall are detected based on the merged crack region.

10. A warehouse wall crack detection device based on three-dimensional laser point cloud data, capable of implementing the warehouse wall crack detection method based on three-dimensional laser point cloud data as described in claim 1, characterized in that, The device includes: The spectrum transformation module is used to acquire the cross-sectional data of the preset target warehouse wall, and to perform spectrum transformation on the cross-sectional data using a preset fast Fourier transform algorithm to obtain the spectrum of the cross-sectional data. The warehouse wall control contour separation module is used to collect three-dimensional laser point cloud data of the target warehouse wall and use a preset fast inverse Fourier transform algorithm to separate the warehouse wall control contour in the three-dimensional laser point cloud data based on the cross-sectional data spectrum. The segmentation threshold calculation module is used to determine the texture distribution features of each cross section in the target warehouse wall based on the control contour elevation of the warehouse wall control contour, and to calculate the segmentation threshold of the cross section based on the texture distribution features. The crack characterization binary image generation module is used to determine the suspected crack points of each cross section according to the segmentation threshold and the control contour elevation, and to combine each of the suspected crack points to obtain a crack characterization binary image. The crack detection module is used to perform region identification on the binary image representing the crack to obtain the crack region, and to detect the cracks in the target warehouse wall based on the crack region using a preset minimum cost spanning tree algorithm.

Images