Wall flatness detection method based on laser area array radar

Abstract

The application relates to the technical field of wall surface flatness detection, and discloses a wall surface flatness detection method based on a laser area array radar, which aims to improve the efficiency and accuracy of wall surface flatness detection. The method uses an area array laser radar system to scan the wall surface, collects reflected laser signals and performs pretreatment. A three-dimensional point cloud model of the wall surface is constructed, and the pretreated signals are accurately matched with the wall surface model by using the spatial resolution capability of the laser radar, so that a three-dimensional laser reflection image of the wall surface is formed. By analyzing the image, machine learning algorithms are used to extract and classify features, automatically identify and label uneven positions, types and degrees of the wall surface. An evaluation model is constructed to evaluate the wall surface flatness, output a flatness score, and provide a comprehensive wall surface flatness detection report in combination with a detection report. The application improves the efficiency and accuracy of wall surface flatness detection, overcomes the limitations of traditional methods, and has a wide application prospect.

Description

Technical Field

[0001] This invention relates to the field of wall flatness detection technology, specifically a wall flatness detection method based on laser array radar. Background Technology

[0002] In fields such as construction engineering, decoration and maintenance, and quality inspection, wall flatness is a key indicator for measuring wall quality. Traditional methods for wall flatness testing mostly rely on manual measurement, which is not only inefficient but also susceptible to subjective factors, making it difficult to guarantee the accuracy and consistency of measurement results. With the advancement of technology, optical measurement technology has gradually been introduced into wall flatness testing to improve measurement efficiency and accuracy. Early optical measurement methods mostly used single-point or line scanning. Compared with manual measurement, these methods improved measurement efficiency to some extent, but still had significant limitations. The measurement range of single-point or line scanning methods is limited, unable to cover the entire wall surface at once, resulting in insufficient data density and difficulty in comprehensively and accurately reflecting the flatness of the wall surface.

[0003] In recent years, lidar technology has been widely used in various fields, such as 3D modeling, terrain mapping, and object recognition, due to its high precision, high efficiency, and non-contact measurement characteristics. In particular, area-array lidar systems can simultaneously acquire spatial information over a large area, offering higher data acquisition speed and density, providing new possibilities for the accurate detection of wall flatness. However, applying area-array lidar to wall flatness detection is not easy and still faces many technical challenges: Effectively acquiring and processing the laser signals reflected from the wall is a key issue. Factors such as wall material, color, and texture can all affect the reflection characteristics of the laser signal; constructing an accurate 3D point cloud model is also a technical challenge. The accuracy of the point cloud model directly affects the reliability of subsequent wall flatness analysis; it is also necessary to accurately map the laser signal onto the wall model to form a 3D laser reflection image, which requires solving the problem of matching the spatial coordinates of the signal and the model. Extracting and classifying features from the 3D laser reflection image to automatically identify the location, type, and degree of wall unevenness is also a crucial step in achieving automated detection. Summary of the Invention

[0004] The purpose of this invention is to provide a wall flatness detection method based on laser array radar to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a wall flatness detection method based on laser array radar, the method comprising:

[0006] The wall surface is scanned using a surface-array lidar system to collect the reflected laser signals from the wall surface;

[0007] The collected reflected laser signals are preprocessed;

[0008] A three-dimensional point cloud model of the wall is constructed, and the spatial resolution capability of the lidar is used to accurately match the pre-processed reflected laser signal with the spatial coordinates of the wall model. The laser signal is then mapped onto the wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall.

[0009] The three-dimensional laser reflection image is analyzed, and machine learning algorithms are used to extract and classify the features in the image, automatically identify and label the location, type and degree of unevenness of the wall, and generate a flatness detection report.

[0010] Construct an evaluation model to assess the flatness of the wall surface and output a flatness score;

[0011] The wall surface flatness is scored using an evaluation model, and the score results are combined with the flatness test report to provide a comprehensive wall surface flatness test report.

[0012] Preferably, the step of mapping the laser signal onto the wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall includes:

[0013] S101: Perform point cloud processing on the preprocessed reflected laser signal. Construct a point cloud matrix P using the point cloud data obtained from multiple scan lines of the lidar. Each row of P corresponds to the point cloud data of one scan line, and each column corresponds to the three-dimensional coordinates of one point.

[0014] S102: Based on the geometric layout of the lidar and the registration principle of point cloud data, calculate the normal vector of the point cloud matrix P to obtain the normal vector matrix N, where the elements of N represent the normal direction of different points; the normal vector is calculated using the following formula:

[0015]

[0016] Among them, P i P i+1 P i+2 For three adjacent point cloud data points, N i Let P be the point i The normal vector;

[0017] S103: Using the normal vector matrix N and the 3D point cloud model of the wall, spatial coordinate matching is performed to map the information in the normal vector to the corresponding spatial position on the wall model, thus obtaining the initial 3D laser reflection image I_0.

[0018] S104: Perform image optimization processing on the initial reflection image I_0 to obtain the optimized reflection image I_e;

[0019] S105: Based on the reflectivity and geometric characteristics of the wall material, set the intensity threshold T_h of the reflected signal and the shape recognition rule R, perform point-by-point analysis on the optimized reflected image I_e', and mark the points whose signal intensity exceeds the threshold T_h and conforms to the shape recognition rule R' as potential uneven areas to obtain the marked reflected image I_m.

[0020] Preferably, the types of wall unevenness identified include protrusions, depressions, cracks, peeling, or deformation of the wall surface.

[0021] Preferably, the image optimization processing step S104 further includes:

[0022] Contrast enhancement is performed on the initial reflection image I_0;

[0023] Edge detection algorithms are used to enhance edge features in images;

[0024] A denoising algorithm is applied to smooth the image.

[0025] Preferably, the step of analyzing the three-dimensional laser reflection image, using machine learning algorithms to extract and classify features in the image, and automatically identifying and labeling the location, type, and degree of unevenness on the wall surface, specifically includes the following steps:

[0026] S201: Construct a deep convolutional neural network model, which includes multiple convolutional layers, pooling layers, fully connected layers, and an output layer. Among them, the convolutional layers are used to extract local features in the reflection image, the pooling layers are used to reduce the feature dimensionality, the fully connected layers are used to integrate features and perform high-level abstraction, and the output layer is used to classify and label the location, type, and degree of unevenness.

[0027] S202: Prepare the training dataset, which contains a large number of labeled LiDAR reflection images of walls. Each image is labeled with the specific location, type and degree of unevenness, which will be used to train the deep convolutional neural network model.

[0028] S203: Input the training dataset into the deep convolutional neural network model, calculate the error between the model's prediction results and the real labels using the backpropagation algorithm and gradient descent method, iteratively adjust the weights and bias parameters in the model until the model's error on the training dataset converges to below the preset threshold.

[0029] S204: Validate the trained deep convolutional neural network model by evaluating its recognition accuracy, recall, and F1 score using an independent validation dataset;

[0030] S205: Input the LiDAR reflection image of the wall to be detected into the trained and validated deep convolutional neural network model. The model automatically extracts, classifies and labels the image, and outputs the specific location, type and degree of unevenness.

[0031] Preferably, the evaluation model is constructed based on the Support Vector Machine (SVM) algorithm.

[0032] Preferably, the process of constructing the evaluation model includes:

[0033] S301: Set the kernel function of the model to the radial basis function RBF;

[0034] S302: Collect training data, which includes wall flatness scores and corresponding reflected laser signal characteristics, including signal intensity, point cloud density and normal vector change rate;

[0035] S303: Input the training data into the SVM model and optimize the model parameters, including the penalty parameter C and the kernel function parameter γ, through the model training algorithm until the model's performance on the validation dataset reaches the preset standard;

[0036] S304: Test the trained SVM model using an independent test dataset to evaluate the model's scoring accuracy and stability.

[0037] Preferably, the construction of the evaluation model further includes a cross-validation step, which uses the K-fold cross-validation method to divide the training data into K subsets, and takes turns using one subset as the validation set and the rest as the training set.

[0038] Preferably, the preprocessing of the acquired reflected laser signal includes: signal denoising, intensity calibration, point cloud registration, and normal estimation.

[0039] Preferably, the method of using an evaluation model to score the flatness of the wall surface and combining the scoring results with a flatness inspection report to provide a comprehensive wall surface flatness inspection report includes:

[0040] The flatness score output by the evaluation model is quantified to form specific scoring indicators, including flatness index, proportion of uneven area and maximum degree of unevenness.

[0041] The quantified scoring indicators are integrated with the information on the location, type, and degree of unevenness in the flatness inspection report to generate a comprehensive report containing detailed scores and unevenness details.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] This invention introduces lidar technology, enabling the lidar system to simultaneously acquire spatial information over a large area with extremely high data acquisition speed. Compared to traditional single-point or line scanning methods, this invention can cover the entire wall surface in one go, significantly improving measurement efficiency. Lidar technology features non-contact and high-precision measurement, accurately capturing subtle changes in the wall surface. By constructing an accurate 3D point cloud model and precisely mapping the laser signal onto the wall model, this invention can comprehensively and accurately reflect the flatness of the wall surface, effectively avoiding subjective errors inherent in manual measurement.

[0044] This invention effectively addresses wall surfaces of varying materials, colors, and textures. By optimizing laser signal acquisition and processing, it ensures accurate measurement results under diverse wall surface conditions, making it highly adaptable and widely applicable in various construction projects, decoration maintenance, and quality inspection scenarios. Highly automated, it reduces labor costs. Utilizing machine learning algorithms to extract and classify features from 3D laser reflection images, it automatically identifies and labels the location, type, and degree of wall unevenness. This automation significantly reduces manual intervention and labor costs. By constructing an evaluation model to assess wall flatness and outputting a flatness score, this invention generates a comprehensive wall flatness inspection report. The report includes measurement results and an intuitive score, allowing users to quickly understand the wall's flatness status. The proposal and implementation of this invention solves the technical challenges in wall flatness inspection. With its widespread application, wall flatness inspection will become more efficient, accurate, and convenient, providing strong support for quality control in construction projects and refined management in decoration maintenance. Attached Figure Description

[0045] Figure 1 This is a schematic diagram illustrating the working principle of the wall flatness detection method based on laser array radar described in this invention.

[0046] Figure 2 A step diagram illustrating the process of mapping laser signals onto a wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall.

[0047] Figure 3 This document outlines the steps for analyzing 3D laser reflection images and using machine learning algorithms to extract and classify features from the images, automatically identifying and labeling the location, type, and degree of unevenness on the wall surface.

[0048] Figure 4 This is a flowchart illustrating the construction process of a wall flatness evaluation model based on the SVM algorithm. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Please see Figure 1-4 This invention provides a technical solution: a method for detecting wall flatness based on laser array radar, the method comprising:

[0051] Wall scanning is performed using an area-array lidar system: This system performs an all-around scan of the target wall, simultaneously emitting and receiving multiple laser beams to quickly acquire the laser signals reflected from the wall. During the scanning process, the relative position of the lidar system to the wall is kept stable to ensure the accuracy of the collected data.

[0052] Preprocessing of the acquired reflected laser signals: Preprocessing operations such as noise reduction and correction are performed on the acquired reflected laser signals to eliminate interference components and improve signal quality. The preprocessing process includes noise filtering and correction of laser beam offset and distortion to ensure the accuracy of subsequent processing.

[0053] Constructing a 3D point cloud model of the wall: Based on the preprocessed reflected laser signal, a 3D point cloud model of the wall is constructed using a point cloud generation algorithm. This model accurately reflects the spatial morphology and structural characteristics of the wall, providing a foundation for subsequent flatness analysis.

[0054] The pre-processed reflected laser signal is precisely matched with the spatial coordinates of the wall model: Utilizing the spatial resolution capability of the lidar, the pre-processed reflected laser signal is precisely matched with the spatial coordinates of the 3D point cloud model of the wall. Then, using a point cloud imaging algorithm, the laser signal is mapped onto the wall model, forming a 3D laser reflection image of the wall.

[0055] Analysis of 3D laser reflection images: Machine learning algorithms are used to analyze 3D laser reflection images and extract feature information, such as uneven features like wall protrusions and depressions. A classification algorithm is then used to automatically identify and label the location, type, and degree of unevenness on the wall surface.

[0056] An evaluation model was constructed to assess wall flatness: Based on the evaluation criteria for wall flatness, an evaluation model was built. This model can comprehensively consider the location, type, and degree of wall unevenness and output a score for wall flatness.

[0057] The wall surface flatness is scored using an evaluation model, and a comprehensive report is generated: The constructed evaluation model is used to score the wall surface flatness, and the scoring results are combined with the flatness inspection report to form a comprehensive wall surface flatness inspection report. This report can comprehensively and accurately reflect the flatness status of the wall surface, providing strong technical support for fields such as building engineering, decoration and maintenance, and quality inspection.

[0058] The present invention will be further described below with reference to Examples 1 to 6:

[0059] Example 1:

[0060] The step of mapping laser signals onto a wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall specifically includes the following implementation methods:

[0061] S101. Point Cloud Processing and Point Cloud Matrix Construction: Point cloud processing is performed on the pre-processed reflected laser signal. The lidar system acquires point cloud data of the wall surface by operating multiple scan lines simultaneously. This point cloud data is organized into a point cloud matrix P, where each row of P corresponds to the point cloud data of one scan line, containing the three-dimensional coordinate information of all points on that scan line; each column corresponds to a specific point, and the elements in the column represent the coordinates of that point in three-dimensional space (such as X, Y, and Z coordinates). In this way, the point cloud matrix P comprehensively reflects the spatial morphology of the wall surface under lidar scanning.

[0062] S102. Normal Vector Calculation: Based on the geometric layout of the lidar and the registration principle of point cloud data, the normal vector of the point cloud matrix P is calculated to obtain the normal vector matrix N, where the elements of N represent the normal direction at different points. The normal vector is calculated using the following formula:

[0063]

[0064] Among them, P i P i+1 P i+2 For three adjacent point cloud data points, N i Let P be the point i The normal vector.

[0065] S103. Spatial Coordinate Matching and Initial Reflection Image Generation: Spatial coordinate matching is performed using the normal vector matrix N and the 3D point cloud model of the wall. The purpose of this step is to map the information in the normal vector to the corresponding spatial position on the wall model, thereby generating the initial 3D laser reflection image I_0.

[0066] In practice, for each normal vector in the normal vector matrix N, its corresponding spatial position in the 3D point cloud model of the wall is found based on the coordinates of its corresponding point in the point cloud matrix P. Then, the information of the normal vector (such as direction, length, etc., which can be selected according to actual needs) is mapped onto that spatial position to form the initial 3D laser reflection image I_0. Each pixel in I_0 corresponds to a specific location on the wall and the normal vector information of that location.

[0067] S104. Image Optimization Processing: The initial reflection image I_0 is subjected to image optimization processing to obtain the optimized reflection image I_e. The purpose of image optimization processing is to improve the image quality and make the features in the image clearer and more accurate.

[0068] Specific optimization methods may include, but are not limited to, denoising, edge enhancement, and contrast adjustment. These processes can eliminate noise and interference in the image, enhance its feature information, and provide a better foundation for subsequent feature extraction and classification.

[0069] S105. Potential Uneven Area Marking: Based on the reflectivity and geometric characteristics of the wall material, set the intensity threshold T_h of the reflected signal and the shape recognition rule R'. Then, perform point-by-point analysis on the optimized reflective image I_e.

[0070] For each pixel in the image, determine whether its signal strength exceeds the threshold T_h and whether it conforms to the shape recognition rule R'. If both conditions are met, mark the point as a potential uneven area. Finally, obtain the marked reflection image I_m. The marked areas in I_m represent the potential uneven areas on the wall.

[0071] Example 2:

[0072] This embodiment describes the steps for preprocessing the acquired reflected laser signal. The specific implementation steps are as follows:

[0073] Signal Denoising: During the scanning process, LiDAR may be affected by various factors such as environmental noise, equipment noise, and wall reflection characteristics, resulting in noise components in the acquired reflected laser signal. To improve the accuracy of subsequent processing, denoising of the original signal is necessary. Specifically, filtering algorithms can be used to denoise the reflected laser signal. For example, common filtering methods such as Gaussian filtering, mean filtering, or median filtering can be used. Appropriate filtering parameters are selected based on the signal characteristics and noise type to smooth the signal, remove noise components, and retain useful signal information.

[0074] Intensity Calibration: The intensity of the reflected laser signal acquired by a lidar may be affected by various factors, such as laser emission power, wall reflectivity, and scanning distance. To accurately reflect the reflectivity of the wall, the signal intensity needs to be calibrated. The intensity calibration process may include: acquiring the reflected laser signal from a standard wall or standard reflector with known reflectivity as a calibration benchmark; establishing an intensity calibration model based on the reflectivity of the standard wall and the acquired signal intensity; and using this model to calibrate the intensity of the actually acquired reflected laser signal to obtain an accurate signal intensity value.

[0075] Point cloud registration: When LiDAR scans a wall, factors such as scanning angle, scanning speed, or equipment movement may cause positional offsets or rotations between point cloud data acquired from different scan lines or different scan cycles. To accurately stitch these point cloud data together to form a complete 3D point cloud model of the wall, point cloud registration is required. The point cloud registration process may include: extracting feature points or feature surfaces from the point cloud data; using feature matching algorithms to find the correspondence between different point cloud data; calculating the transformation matrix based on the correspondence, and performing rotation and translation transformations on the point cloud data to achieve accurate point cloud registration.

[0076] Normal estimation: For subsequent spatial coordinate matching and image generation, it is necessary to estimate the normal vector of the wall point cloud. The normal vector is an important feature describing the orientation of the point cloud surface and is of great significance for wall flatness detection. The normal estimation process may include: for each point in the point cloud, selecting several points in its neighborhood as reference points; then, using the coordinate information of these reference points, estimating the normal vector of that point through geometric calculations or statistical methods; finally, normalizing the estimated normal vector to obtain the unit normal vector.

[0077] Example 3:

[0078] The image optimization processing step S104 further includes the following detailed implementation:

[0079] The initial reflected image I_0 undergoes contrast enhancement processing to improve its visual effect and make details more clearly discernible. Contrast enhancement can be achieved by adjusting the image's grayscale or color values, making bright areas brighter and dark areas darker, thereby enhancing the image's sense of depth and three-dimensionality. Specifically, methods such as histogram equalization, linear stretching, or non-linear transformations can be used to enhance image contrast. For example, histogram equalization calculates the image's grayscale histogram and then redistributes the grayscale values ​​to ensure a uniform distribution across the entire range, thus enhancing image contrast.

[0080] Edge detection algorithms are used to enhance edge features in images, highlighting edge information and facilitating subsequent feature extraction and classification. Edges are locations in an image where grayscale values ​​change drastically, typically corresponding to the contours or boundaries of objects. Specific edge detection algorithms can include the Sobel operator, Canny operator, and Prewitt operator. These algorithms detect the position and orientation of edges by calculating the gradient value or second derivative of each pixel in the image. The detected edges can then be thinned or connected to obtain more complete and accurate edge features.

[0081] Denoising algorithms are applied to smooth images, eliminating noise and interference to improve image quality and reliability. Noise consists of unwanted random variations or fluctuations in an image that can obscure useful information or produce misleading results. Specific denoising algorithms include mean filtering, median filtering, Gaussian filtering, bilateral filtering, etc. These algorithms smooth the image and reduce noise by calculating the average, median, or weighted average of each pixel and its neighborhood pixels. When selecting a denoising algorithm, it is necessary to choose an appropriate algorithm and parameters based on the characteristics of the image and the type of noise to achieve the best denoising effect.

[0082] Example 4:

[0083] In this invention, the specific steps for analyzing three-dimensional laser reflection images and using machine learning algorithms to extract and classify features from the images to automatically identify and label the location, type, and degree of unevenness on the wall surface include:

[0084] S201. Constructing a deep convolutional neural network model:

[0085] A deep convolutional neural network model is constructed, consisting of multiple convolutional layers, pooling layers, fully connected layers, and an output layer. The convolutional layers extract local features from the reflection image, such as edges and textures, converting the input image into a feature map through convolution operations. Pooling layers reduce the dimensionality of the feature maps, decreasing computation while preserving important features. The fully connected layers integrate the features output from the pooling layers and perform high-level abstraction, forming a holistic understanding of the image. The output layer, based on the output of the fully connected layers, classifies and labels the location, type, and degree of unevenness.

[0086] In practice, appropriate parameters such as kernel size, stride, and padding can be selected to construct convolutional layers, and appropriate pooling window size and stride can be selected to construct pooling layers. The number of neurons in the fully connected layer can be adjusted according to the specific task, while the number of neurons in the output layer corresponds to the number of classification and labeling categories.

[0087] S202. Prepare the training dataset:

[0088] Prepare a training dataset containing a large number of labeled LiDAR reflection images of walls. Each image is labeled with the specific location, type, and degree of unevenness, such as cracks, bulges, depressions, etc., as well as the severity of the unevenness. This labeled information will serve as supervision signals for model training, guiding the model to learn how to identify and label the location, type, and degree of unevenness on the wall.

[0089] The training dataset should cover a variety of wall surface conditions, including different materials, lighting conditions, and types and degrees of unevenness, to ensure that the model has generalization ability and can adapt to wall flatness detection tasks in different scenarios.

[0090] S203, Training a deep convolutional neural network model:

[0091] The training dataset is fed into a deep convolutional neural network model, and the error between the model's predictions and the ground truth labels is calculated using backpropagation and gradient descent. Then, the weights and biases in the model are iteratively adjusted to minimize the error until the model's error on the training dataset converges to below a preset threshold.

[0092] During training, appropriate optimizers (such as Adam, SGD, etc.) and learning rates can be selected to accelerate convergence, while regularization techniques (such as L2 regularization, Dropout, etc.) can be used to prevent overfitting. Furthermore, data augmentation techniques (such as rotation, scaling, cropping, etc.) can be used to expand the training dataset and improve the model's robustness.

[0093] S204. Validate the deep convolutional neural network model:

[0094] The trained deep convolutional neural network model was validated using an independent validation dataset to evaluate its accuracy, recall, and F1 score. The validation dataset should contain wall-mounted LiDAR reflection images that are different from the training dataset but have a similar distribution to ensure the objectivity and accuracy of the evaluation results.

[0095] By calculating various metrics of the model on the validation dataset, we can evaluate whether the model's performance meets the requirements. If the model performance is unsatisfactory, we can return to step S201 to adjust the model structure or step S203 to optimize the training process until the model performance reaches the expected level.

[0096] S205. Applying a deep convolutional neural network model to detect wall flatness:

[0097] The LiDAR reflection image of the wall surface to be inspected is input into a trained and validated deep convolutional neural network model. The model automatically extracts, classifies, and labels features from the image, outputting information on the specific location, type, and degree of unevenness. This information can be used to generate a wall surface flatness inspection report, providing a basis for subsequent maintenance and renovation.

[0098] In practical applications, deep convolutional neural network models can be integrated into wall flatness detection systems to achieve automated and intelligent detection processes. Simultaneously, new data can be continuously collected to update and optimize the model, adapting to ever-changing wall conditions and detection needs.

[0099] Example 5:

[0100] The evaluation model is constructed based on the Support Vector Machine (SVM) algorithm. The detailed construction process of this model includes:

[0101] S301. Define the kernel function of the model: Select the radial basis function (RBF) as the kernel function of the SVM model. The RBF kernel function is a commonly used nonlinear kernel function that can map data in the original space to a high-dimensional space, thereby handling nonlinear problems.

[0102] S302. Collect training data: This data includes wall flatness scores and corresponding reflected laser signal characteristics. Wall flatness scores can be continuous or discrete values ​​obtained from expert evaluation or standard measurements. Reflected laser signal characteristics include signal intensity, point cloud density, and rate of change of the normal vector, which reflect the geometric and reflective properties of the wall. Ensure the training data is representative and diverse, covering different types of walls, different flatness conditions, and different environmental conditions.

[0103] S303. Model Training: Input the training data into the SVM model and optimize the model parameters through the model training algorithm. The parameters of the SVM model include the penalty parameter C and the kernel function parameter γ. The penalty parameter C controls the degree of punishment for classification errors. A larger C value will cause the model to pay more attention to the accuracy of the training data, but may lead to overfitting; a smaller C value will cause the model to generalize more, but may lead to underfitting.

[0104] The model is trained using the training dataset, and the optimal combination of C and γ parameters is selected through cross-validation or other methods. Specifically, methods such as grid search can be used to traverse within a preset parameter range and select the parameter combination that best performs the model on the validation dataset.

[0105] Furthermore, K-fold cross-validation is used to further validate the model's performance. The training data is divided into K subsets, with one subset used as the validation set and the rest as the training set in turn. Through K rounds of cross-validation, the model's stability and generalization ability on different datasets can be evaluated.

[0106] S304. Model Testing: Test the trained SVM model using an independent test dataset to evaluate its scoring accuracy and stability. The test dataset should contain wall flatness scores and reflected laser signal features that are different from the training dataset but have a similar distribution.

[0107] By calculating the model's scoring accuracy and stability on the test dataset, we can evaluate whether the model's performance meets the requirements. If the model's performance is unsatisfactory, we can return to step S301 to adjust the kernel function or step S303 to optimize the model parameters until the model's performance reaches the expected level.

[0108] Example 6:

[0109] This invention utilizes an evaluation model to score the flatness of a wall surface and combines the scoring results with a flatness inspection report to provide a comprehensive wall surface flatness inspection report. The method specifically includes the following steps:

[0110] ① The flatness score output by the evaluation model is quantified to form specific scoring indicators. These scoring indicators include, but are not limited to:

[0111] Smoothness Index: This is an indicator that comprehensively reflects the overall smoothness of the wall surface. It can be obtained by normalizing or standardizing the original score output by the evaluation model. The higher the smoothness index, the smoother the wall surface; conversely, the lower the smoothness index, the more uneven areas there are on the wall surface.

[0112] Uneven area ratio: This refers to the proportion of uneven areas to the total wall area. It can be obtained by calculating the area of ​​the uneven areas identified by the evaluation model and comparing it with the area of ​​the entire wall. A higher uneven area ratio indicates more uneven areas in the wall.

[0113] Maximum Unevenness: This refers to the score or measurement value of the area with the most severe unevenness on the wall surface. The maximum unevenness index can be selected by ranking the unevenness scores output by the evaluation model and choosing the highest value.

[0114] ② The quantified scoring indicators are then integrated with the information on the location, type, and degree of unevenness in the flatness inspection report. The specific integration method is as follows:

[0115] The flatness inspection report clearly lists scoring indicators such as the flatness index, the proportion of uneven areas, and the maximum degree of unevenness, to visually demonstrate the overall flatness of the wall surface. Simultaneously, the report records in detail the location, type, and severity of unevenness identified by the assessment model. This information can include the specific location of the uneven areas (e.g., coordinates, area number), the type of unevenness (e.g., cracks, protrusions, depressions), and the degree of unevenness (e.g., minor, moderate, severe). By combining the scoring indicators with the unevenness details, a comprehensive report containing detailed scores and unevenness information is generated. This report not only comprehensively reflects the flatness of the wall surface but also provides specific unevenness information, offering a strong basis for subsequent repairs and renovations.

[0116] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0117] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting wall flatness based on laser array radar, characterized in that, The method includes: A surface-array lidar system is used to scan the wall and collect the reflected laser signals from the wall. The collected reflected laser signals are preprocessed; A three-dimensional point cloud model of the wall is constructed, and the spatial resolution capability of the lidar is used to accurately match the pre-processed reflected laser signal with the spatial coordinates of the wall model. The laser signal is then mapped onto the wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall. The three-dimensional laser reflection image is analyzed, and machine learning algorithms are used to extract and classify the features in the image, automatically identify and label the location, type and degree of unevenness of the wall, and generate a flatness detection report. Construct an evaluation model to assess the flatness of the wall surface and output a flatness score; The wall surface flatness is scored using an evaluation model, and the scoring results are combined with the flatness test report to provide a comprehensive wall surface flatness test report. The step of mapping laser signals onto a wall model using a point cloud imaging algorithm to form a three-dimensional laser reflection image of the wall includes: S101: Perform point cloud processing on the preprocessed reflected laser signal. Construct a point cloud matrix P using the point cloud data obtained from multiple scan lines of the lidar. Each row of P corresponds to the point cloud data of one scan line, and each column corresponds to the three-dimensional coordinates of one point. S102: Based on the geometric layout of the lidar and the registration principle of point cloud data, calculate the normal vector of the point cloud matrix P to obtain the normal vector matrix N, where the elements of N represent the normal direction of different points; the normal vector is calculated using the following formula: in, For three adjacent point cloud data points, For point The normal vector; S103: Using the normal vector matrix N and the 3D point cloud model of the wall, spatial coordinate matching is performed to map the information in the normal vector to the corresponding spatial position on the wall model, thus obtaining the initial 3D laser reflection image I_0. S104: Perform image optimization processing on the initial reflection image I_0 to obtain the optimized reflection image I_e; S105: Based on the reflectivity and geometric characteristics of the wall material, set the intensity threshold T_h of the reflected signal and the shape recognition rule R, perform point-by-point analysis on the optimized reflected image I_e', and mark the points whose signal intensity exceeds the threshold T_h and conforms to the shape recognition rule R' as potential uneven areas to obtain the marked reflected image I_m.

2. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The types of wall irregularities that can be identified include bumps, depressions, cracks, peeling, or deformation of the wall surface.

3. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The image optimization processing step S104 further includes: Contrast enhancement is performed on the initial reflection image I_0; Edge detection algorithms are used to enhance edge features in images; A denoising algorithm is applied to smooth the image.

4. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The analysis of the three-dimensional laser reflection image, using machine learning algorithms to extract and classify features from the image, and automatically identifying and labeling the location, type, and degree of unevenness on the wall surface, includes the following specific steps: S201: Construct a deep convolutional neural network model, which includes multiple convolutional layers, pooling layers, fully connected layers, and an output layer. Among them, the convolutional layers are used to extract local features in the reflection image, the pooling layers are used to reduce the feature dimensionality, the fully connected layers are used to integrate features and perform high-level abstraction, and the output layer is used to classify and label the location, type, and degree of unevenness. S202: Prepare the training dataset, which contains a large number of labeled LiDAR reflection images of walls. Each image is labeled with the specific location, type and degree of unevenness, which will be used to train the deep convolutional neural network model. S203: Input the training dataset into the deep convolutional neural network model, calculate the error between the model's prediction results and the real labels using the backpropagation algorithm and gradient descent method, iteratively adjust the weights and bias parameters in the model until the model's error on the training dataset converges to below the preset threshold. S204: Validate the trained deep convolutional neural network model by evaluating its recognition accuracy, recall, and F1 score using an independent validation dataset; S205: Input the LiDAR reflection image of the wall to be detected into the trained and validated deep convolutional neural network model. The model automatically extracts, classifies and labels the image, and outputs the specific location, type and degree of unevenness.

5. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The evaluation model is constructed based on the Support Vector Machine (SVM) algorithm.

6. The wall flatness detection method based on laser array radar according to claim 5, characterized in that, The process of constructing the evaluation model includes: S301: Set the kernel function of the model to the radial basis function RBF; S302: Collect training data, which includes wall flatness scores and corresponding reflected laser signal characteristics, including signal intensity, point cloud density and normal vector change rate; S303: Input the training data into the SVM model and optimize the model parameters, including the penalty parameter C and the kernel function parameter γ, through the model training algorithm until the model's performance on the validation dataset reaches the preset standard; S304: Test the trained SVM model using an independent test dataset to evaluate the model's scoring accuracy and stability.

7. The wall flatness detection method based on laser array radar according to claim 6, characterized in that, The construction of the evaluation model also includes a cross-validation step, which uses the K-fold cross-validation method to divide the training data into K subsets, and alternately uses one subset as the validation set and the rest as the training set.

8. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The preprocessing of the acquired reflected laser signal includes: signal denoising, intensity calibration, point cloud registration, and normal estimation.

9. The wall flatness detection method based on laser array radar according to claim 1, characterized in that, The method for using an evaluation model to score the flatness of a wall surface, and combining the scoring results with a flatness inspection report to provide a comprehensive wall surface flatness inspection report includes: The flatness score output by the evaluation model is quantified to form specific scoring indicators, including flatness index, proportion of uneven area and maximum degree of unevenness. The quantified scoring indicators are integrated with the information on the location, type, and degree of unevenness in the flatness inspection report to generate a comprehensive report containing detailed scores and unevenness details.

Images