An intelligent management system and method for wallboard installation

Abstract

The application aims to provide an intelligent management system and method for wallboard installation, which comprises the following steps: acquiring a wall image; optimizing a wallboard installation management model; inputting the wall image into the wallboard installation management model, and outputting a wallboard installation result from the wallboard installation management model. The application has wall position detection capability and wallboard layout prediction capability, and comprehensively covers the wallboard installation of the whole house. The application combines various detection results, intelligently evaluates the wallboard installation management scheme, improves the accuracy and reliability of the wallboard installation, and improves the precision of the wallboard layout through a genetic algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and particularly relates to an intelligent management system and method for wallboard installation. BACKGROUND

[0002] Image recognition is an important field of artificial intelligence that uses computers to process, analyze, and understand images to recognize various patterns of objects and subjects. Image recognition technology automatically identifies and classifies objects, scenes, and other information in images by analyzing their features and patterns. Image recognition, also known as computer vision, is a technology that uses computers and artificial intelligence to analyze and understand images. It not only recognizes objects and scenes in two-dimensional images, but also processes three-dimensional objects and environmental perception. Image recognition technology has a wide range of applications in various industries, including manufacturing, healthcare, retail, agriculture, and security monitoring. In manufacturing, image recognition is used for automated quality control; in healthcare, it helps doctors more accurately detect diseases; in retail, image recognition can provide personalized shopping experiences; in agriculture, it helps identify crop problems and take action; and in security monitoring, image recognition is used to monitor and detect anomalies.

[0003] Traditional image processing methods mainly rely on edge detection, color segmentation, and other traditional computer vision and image processing techniques. Machine learning methods based on statistical machine learning algorithms have begun to be applied to image recognition. The emergence of deep learning methods such as convolutional neural networks (CNN) has greatly promoted the development of image recognition, enabling end-to-end image recognition tasks. Pre-trained models and transfer learning: by pre-training models on large-scale datasets and then fine-tuning them on specific tasks, model efficiency is improved. In recent years, image recognition has combined with other technologies such as natural language processing (NLP) and augmented reality (AR), driving innovative applications.

[0004] Traditional wallboard installation process requires the use of measuring tools such as tape measures or levels to measure wall dimensions and positions, and mark the installation position of the wallboard on the wall. This method has low installation efficiency and low installation accuracy, so artificial intelligence can be used to assist in the installation of wallboards. Currently, when artificial intelligence identifies the installation position of wallboards, most wallboards are white, but white itself does not have obvious color contrast, so it is difficult to determine the range and type of the installation position of the wallboard during image detection, making it difficult to accurately install the wallboard. SUMMARY

[0005] The application aims to provide an intelligent management system and method for wallboard installation, which has wall surface position detection capability and wallboard layout prediction capability, covers wallboard installation of the whole house, combines various detection results, intelligently evaluates wallboard installation management scheme, improves the accuracy and reliability of wallboard installation, and improves the precision of wallboard layout through genetic algorithm.

[0006] An intelligent management method for wallboard installation comprises:

[0007] Obtaining a wall surface image;

[0008] Optimizing a wallboard installation management model;

[0009] Inputting the wall surface image into the wallboard installation management model, and outputting a wallboard installation result from the wallboard installation management model.

[0010] Preferably, after obtaining the wall surface image, the method further comprises preprocessing the wall surface image, specifically:

[0011] Grayscale processing the wall surface image;

[0012] Denoising the grayscale wall surface image;

[0013] Enhancing the denoised wall surface image;

[0014] Standardizing the size of the enhanced wall surface image.

[0015] Preferably, the optimization of the wallboard installation management model comprises:

[0016] Replacing the architecture of the wallboard installation management model with ResNet50 feature extraction;

[0017] Adding RPN and RoI pooling detection layers to the wallboard installation management model;

[0018] Optimizing the loss function of the wallboard installation management model;

[0019] Adding a genetic algorithm to the wallboard installation management model.

[0020] Preferably, after denoising the grayscale wall surface image, the method further comprises judging whether the denoising effect meets the standard, specifically:

[0021] The denoising effect is represented as:

[0022]

[0023] wherein f is the original image, g is the denoised image, u f is the average value of the original image, and u gthe average value of the denoised image, σ (f，g) is the covariance of the original image and the denoised image, σ f is the variance of the original image, σ g the variance of the denoised image, A1 and A2 are constants.

[0024] Preferably, the architecture of the wallboard installation management model is replaced with ResNet50 feature extraction, including:

[0025] ResNet50 is used for feature extraction of wall surface images, denoted as:

[0026]

[0027] where w ij is the weight of the i-th row and j-th column of the convolution kernel, x ij is the element of the i-th row and j-th column of the input feature map, b is the bias term, and the size of the convolution kernel is m x n.

[0028] Preferably, the loss function of the wallboard installation management model is optimized, including:

[0029]

[0030] where p i is the probability of the i-th anchor frame predicting the target, is the true value of the i-th anchor frame, which is 1 when it is a positive sample, otherwise it is 0, L1 is the classification loss function, L2 is the regression loss function, L t is the total loss function, N1 is the number of training anchor frames, N2 is the number of anchor frames that need to be regressed, α is the balance coefficient, smooth is the smoothing function, t i is the offset parameter of the predicted frame and the corresponding anchor frame, is the offset parameter of the real target frame and the corresponding anchor frame.

[0031] Preferably, the genetic algorithm is added to the wallboard installation management model, including:

[0032] The wall surface data and the wallboard size are input into the genetic algorithm as a training set;

[0033] The genetic algorithm is trained;

[0034] The trained genetic algorithm is used to calculate the installation layout and the installation quantity of the wallboard according to the wall surface data and the wallboard size.

[0035] A wallboard defect intelligent detection system based on image recognition, comprising:

[0036] An image acquisition module for acquiring wall surface images;

[0037] a model optimization module, configured to optimize a wallboard installation management model;

[0038] an image processing module, configured to input the wall surface image into the wallboard installation management model, and output a wallboard installation result from the wallboard installation management model.

[0039] An electronic device, comprising a chip, a processor and a memory, the memory being configured to store computer program code, the computer program code comprising computer instructions, the electronic device being configured to execute an intelligent management method for wallboard installation when the chip executes the computer instructions.

[0040] A computer readable storage medium, the computer readable storage medium storing a computer program, the computer program comprising program instructions, the program instructions being configured to cause a processor to execute an intelligent management method for wallboard installation when the program instructions are executed by the processor.

[0041] The present application has the following beneficial effects: 1. The wall surface image is preprocessed in the present application, the installation position in the preprocessed wall surface image is easier to be detected, and the machine vision can better identify the wallboard installation position 2. The wallboard installation management model is optimized and improved in the present application, the model is more sensitive to the wall surface installation position, the feature extraction network is replaced so as to better extract the features related to the wallboard installation position, and the loss function is optimized so that the detection accuracy of the model is higher. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without creative labor.

[0044] Figure 1 A flowchart of an intelligent management method for wallboard installation according to the present application;

[0045] Figure 2 A flowchart of a wallboard installation management model optimization according to the present application;

[0046] Figure 3 A hardware structure schematic diagram of an electronic device according to the present application. DETAILED DESCRIPTION

[0047] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0048] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative positional relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications will also change accordingly.

[0049] In addition, the description of "first", "second" and the like in the present application is only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of those skilled in the art, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor within the scope of protection required by the present application.

[0050] The traditional wallboard installation process needs to use measuring tools (such as tape measure or level) to measure the size and position of the wall surface, and mark the installation position of the wallboard on the wall surface. The installation efficiency of this method is very low, and the installation accuracy is also not high, so artificial intelligence can be used to intelligently assist the installation of the wallboard. At present, when artificial intelligence identifies the installation position of the wallboard, since most of the wallboards are white, but white itself has no obvious color contrast, so it is difficult to determine the range and type of the installation position of the wallboard which is not very obvious during image detection, and it is difficult to accurately install the wallboard.

[0051] The present application pre-processes the wall surface image, and the installation position in the pre-processed wall surface image is easier to be detected, which can help machine vision to better identify the wallboard installation position. The present application optimizes and improves the wallboard installation management model, so that the model is more sensitive to the wall surface installation position. The feature extraction network is replaced so as to better extract the features related to the wallboard installation position. The loss function is optimized so that the detection accuracy of the model is higher.

[0052] Embodiment 1

[0053] An intelligent management method for wallboard installation, with reference to Figure 1 , comprising:

[0054] S100, acquire a wall image;

[0055] The laser radar scans the wall surface by emitting a laser beam and measuring the time it takes for the beam to reflect back, determining the distance and position of the target object, and thus constructing a three-dimensional point cloud map of the surrounding environment. The working principle of the laser radar can be summarized as "emission-reflection-reception" three steps. It performs continuous scanning in the specified area through high-speed rotation or electronic scanning, and when it detects a target entering the set warning area, it can quickly locate the target direction and track it in real time. When identifying the side wall, the laser radar will scan the side of the wall, and then calculate the distance and position of the wall according to the reflected laser signal. Through the combination of multiple scanning points, the laser radar can construct the three-dimensional shape of the wall, and thus achieve accurate identification of the side wall.

[0056] The point cloud data of the wall surface can be obtained by laser radar scanning, which can be used in three-dimensional modeling, scene reconstruction, robot navigation, virtual reality and augmented reality, etc. Through the point cloud data, a realistic three-dimensional scene model can be generated. In the embodiment of the present application, the point cloud data is used to calculate the parameters of the wall, including the height of the wall, the width of the wall and the position of the wall.

[0057] S200, optimize the wallboard installation management model;

[0058] The main function of the wallboard installation management model is to calculate the installation scheme of the wallboard according to the obtained wall surface parameters. The content of the wallboard installation scheme includes: installing the wallboard at the specified position of the specified wall surface according to the specified angle. The wallboard installation management model first needs to process the wall image, capture the specified position of the wall image, then match the corresponding wallboard that needs to be installed, and then control the wallboard to install the wallboard according to the specified angle. Then output the installation scheme to the mechanical control module, and the mechanical control module installs the wallboard according to the installation scheme. The robot system developed by artificial intelligence technology is used for automatic installation of wallboard. The mechanical control module includes robots, sensors, control systems and related software, which can automatically complete the operations of wallboard picking, clamping, lifting, turning and installing, etc.

[0059] S300, input the wall image into the wallboard installation management model, and the wallboard installation management model outputs the wallboard installation result.

[0060] In the embodiment of the present application, the wallboard installation management model can accurately identify the specific position of the wall surface that needs to install the wallboard, and can match the type of wallboard that needs to be installed in the design scheme. In the wallboard installation management scheme, first, each wall surface that needs to install the wallboard is labeled, then the wallboard to be installed is labeled, and then the wall surface is identified. After identifying the wall surface that needs to install the wallboard in the design scheme, the installation position of the wall surface is found. After determining the installation position, the control signal is output to the mechanical control module to control the intelligent installation of the wallboard. The present application does not need to use manual control, can intelligently install the wallboard according to the input design scheme, has high installation efficiency, accurate installation position, avoids installation defects caused by manual calculation errors, and can also install multiple wallboards at the same time, which can greatly improve the wallboard installation efficiency.

[0061] Preferably, after obtaining the wall surface image, S100 further comprises S110, preprocessing the wall surface image, specifically:

[0062] S111, graying the wall surface image;

[0063] Because the wall surface is mostly pure color or has a small amount of pattern, and the color and pattern of the wall surface have little effect on the installation position of the wallboard, the wall surface image needs to be graying to reduce the influence of unnecessary colors, and also to simplify image processing, improve calculation efficiency, reduce data volume, and enhance the recognizability of image features. Each pixel in a color image is determined by three components: red (R), green (G), and blue (B), which means that the data volume of a color image is relatively large and complex to process. Each pixel in a grayscale image has only one grayscale value, which simplifies the representation of the image and makes the processing process more efficient. Each pixel in a color image can have 16777216 color variations (256 cubed), while each pixel in a grayscale image can have only 256 color variations (0-255). Therefore, converting a color image to a grayscale image can significantly reduce the amount of calculation and improve the processing speed. In a grayscale image, the grayscale value of each pixel reflects the brightness information of the point, which helps to better analyze and identify the features of the wall surface image in image processing and computer vision tasks.

[0064] S112, denoising the graying wall surface image;

[0065] The main role of image denoising is to improve the quality of the image, making it more suitable for subsequent computer vision processing and analysis. Image denoising technology can significantly improve the clarity and detail of the wall image by removing noise in the wall image, thereby improving the overall visual effect of the wall image. By removing random noise in the wall image, such as light interference, sensor noise, etc., image denoising technology can make the wall image clearer, reduce blurring and distortion, and improve the overall quality of the image. In the wallboard installation management model, denoising can reduce the interference of noise on wall image recognition and classification, and improve the performance and accuracy of the system.

[0066] S113, enhancing the denoised wall image;

[0067] The main role of image enhancement is to improve the visual effect and quality of the image, making it more suitable for specific application requirements. By enhancing the overall or local characteristics of the image, image enhancement technology can make the originally unclear image clear, enhance certain features of interest, enlarge the differences between different object features in the image, and suppress uninteresting features, thereby improving the quality and information content of the image. In the embodiments of the present application, the information such as the texture of the wall can be enhanced to strengthen the identification of the installation position of the wall installation management model.

[0068] S114, size standardization of the enhanced wall image.

[0069] By standardizing the image size, the size difference caused by image acquisition equipment and environment, etc. can be eliminated, making the wall image more consistent and reliable in the identification process, thereby improving the accuracy of identification. The standardized image size can reduce the computational complexity, making the image processing algorithm more efficient. For example, in processing a large number of wall images, uniform size can simplify the processing flow and reduce the consumption of computing resources. Standardization processing can eliminate errors caused by different sizes, enhance the robustness of the wallboard installation management model, and maintain a high recognition rate when facing different sizes of wall image input.

[0070] Preferably, S200, optimizing the wallboard installation management model comprises:

[0071] S210, replacing the architecture of the wallboard installation management model with ResNet50 feature extraction;

[0072] ResNet-50 is a deep convolutional neural network, and its network structure includes the following parts: input layer: usually a 224x224 pixel RGB image with 3 channels. Initialization convolution layer: 7x7 convolution kernel, 64 filters, stride 2. Batch normalization layer: normalization is performed to improve the stability of training. ReLU activation layer: introduces nonlinearity. Max pooling layer: 3x3 pooling kernel, stride 2. Residual block: there are 16 residual blocks in total, divided into four stages, and the number of residual blocks and filters in each stage is different. Global average pooling layer: the spatial dimension of each channel of the feature map is compressed into a single value through average pooling, generating a global feature vector. Fully connected layer: used for the final classification task, followed by a Softmax layer for classification decision.

[0073] S220, adding RPN and RoI pooling detection layers to the wallboard installation management model;

[0074] RPN (Region Proposal Network) detection layer is a key component in target detection algorithms, mainly used to generate candidate target regions. RPN generates multiple candidate boxes (or called candidate regions) by sliding windows on the feature map, and scores and coordinates regress each candidate box to filter out candidate boxes that may contain targets.

[0075] RoI pooling detection layer is a technique used in target detection algorithms to handle different sizes of target regions. The main role of RoI pooling (Region of Interest pooling) is to normalize different sizes of candidate regions (RoI) to fixed size outputs for subsequent processing. RoI pooling layer divides the input candidate regions into multiple sub-regions and performs maximum pooling (Max Pooling) operation on each sub-region to obtain fixed size output. The specific process is as follows: input: the input of RoI pooling layer includes feature map and a series of candidate regions. The feature map is the feature extracted by convolutional neural network (CNN), and the RoI is the mapping of the candidate box generated by RPN on the feature map. Divide sub-regions: divide each RoI into multiple sub-regions, and the size of each sub-region is determined according to the output size. Pooling operation: maximum pooling operation is performed on each sub-region to obtain fixed size output.

[0076] S230, optimizing the loss function of the wallboard installation management model;

[0077] The role of the loss function in machine learning is mainly to quantify the difference between the model's predicted value and the actual value, and to optimize the model parameters by minimizing this difference, thereby improving the model's prediction accuracy. During the training process, the loss function serves as the objective function, guiding the adjustment of model parameters to minimize prediction errors. The loss function calculates a numerical value to represent the accuracy or error size of the model's prediction, and the model's goal is to minimize the value of the loss function by adjusting its parameters, thereby improving the accuracy of the prediction.

[0078] S240, add a genetic algorithm to the wallboard installation management model.

[0079] In the embodiments of the present application, the genetic algorithm is used to find the optimal wallboard layout and quantity. Through continuous iteration of the genetic algorithm, the most suitable installation scheme is found among the reasonable wallboard installation schemes.

[0080] Preferably, after the grayscale wall image is denoised, it is further judged whether the denoising effect meets the standard, specifically:

[0081] The denoising effect is represented as:

[0082]

[0083] Where f is the original image, g is the denoised image, uf is the average value of the original image, u g The average value of the denoised image, σ (f，g) is the covariance of the original image and the denoised image, σ f is the variance of the original image, σ g The variance of the denoised image, A1 and A2 are constants.

[0084] Because the wall surface is relatively smooth, there are few feature points that can be extracted, and if there is noise, it will have a great impact on the wall recognition result. Since the noise of the wall image has a great impact on the subsequent wallboard installation scheme, the wall image denoising effect needs to be evaluated. Only the wall image that has passed the denoising effect evaluation can enter the processing of the wallboard installation management model.

[0085] Denoising technology can reduce granular spots and discoloration spots in the image, improving the clarity and realism of the image. The denoised wall image retains important features in the image, such as edges, textures, etc., making subsequent image processing and analysis more accurate and effective, and accurately extracting feature information of the wall, improving the prediction accuracy.

[0086] Preferably, S210, replacing the architecture of the wallboard installation management model with ResNet50 feature extraction includes:

[0087] ResNet50 is used to extract features from wall images, which can be expressed as:

[0088]

[0089] Among them, w ij is the weight of the convolution kernel in row i and column j, x ij is the element in the i-th row and j-th column of the input feature map, b is the bias term, and the size of the convolution kernel is m×n.

[0090] Feature extraction can reduce the dimensionality of raw data, remove redundant and irrelevant information, and retain only features that are useful for learning, thereby improving the learning efficiency and accuracy of the model. Feature extraction can help the model learn the inherent laws in the data and improve the generalization ability of the model.

[0091] In this embodiment of the present invention, wall image feature extraction can reduce the dimensionality of the data, simplifying the subsequent modeling and analysis process, thereby improving processing speed and efficiency. By extracting relevant features, the accuracy and performance of the machine learning model can be improved, overfitting can be reduced, and the data after feature extraction is smaller and easier to process, which can accelerate the processing speed of the machine learning algorithm.

[0092] Preferably, in S230, optimizing the loss function of the wall panel installation management model includes:

[0093]

[0094] Among them, pi is the probability that the i-th anchor box is predicted to be the target, is the true value of the i-th anchor box. When it is a positive sample, the value is 1, otherwise it is 0. L1 is the classification loss function, L2 is the regression loss function, and L t is the total loss function, N1 is the number of training anchor boxes, N2 is the number of anchor boxes that need to be regressed, α is the balance coefficient, smooth is the smoothing function, t i is the offset parameter between the predicted box and the corresponding anchor box, is the offset parameter between the real target box and the corresponding anchor box.

[0095] In the embodiment of the present application, the total loss function is composed of two loss functions, the regression loss function refers to a function used to measure the difference between the predicted value and the actual observed value in a regression problem. The regression problem aims to predict a continuous value output through a set of input features, so the design of the loss function needs to be able to reflect the error between the predicted value and the true value. The classification loss function is used to evaluate the performance of the classification model, usually using the cross-entropy loss function, which is used to measure the difference between the predicted value and the true value of the classification model. In machine learning and deep learning, the loss function is used as the target of optimization, by minimizing the value of the loss function to adjust the model parameters, so as to improve the prediction accuracy of the model. The commonly used optimization algorithm such as gradient descent is to update the model parameters by calculating the gradient of the loss function. The loss function can help monitor the overfitting of the model during the training process. By observing the loss on the training set and the validation set, it can be found early whether the model is overfitting, and appropriate measures such as regularization can be taken. The loss function can introduce weights to balance the importance between different samples or objectives, so as to more accurately measure the performance of the model. The present application uses a combination of multiple loss functions to evaluate the wallboard installation management model, and the loss function has the advantages of fast convergence and strong robustness, and can cope with multiple scenarios, wide application range and strong adaptability.

[0096] Preferably, S240, the genetic algorithm is added to the wallboard installation management model, including:

[0097] S241, input the wall surface data and wallboard size into the genetic algorithm as a training set;

[0098] Genetic algorithm is an optimization search method based on the principles of natural selection and genetics. Genetic algorithm simulates the process of Darwin's natural selection and genetic inheritance in biological evolution, iteratively updates the solution to the problem, and searches for the optimal solution or approximate optimal solution. The basic idea of genetic algorithm is as follows: map the problem into a mathematical problem, that is, establish a mathematical model. Initialize a population containing multiple individuals, each individual representing a solution. Perform selection operation, evaluate individuals according to fitness function, and select excellent individuals for breeding. Perform crossover operation, randomly select two individuals for chromosome crossover to produce new offspring. Perform mutation operation on the offspring's chromosome. Update the optimal solution, add excellent offspring to the population, and eliminate some poor individuals.

[0099] S242, train the genetic algorithm;

[0100] The main role of genetic algorithm training is to solve complex optimization problems. Genetic algorithm simulates the principles of natural selection and genetics, and gradually optimizes the solution of the problem through selection, crossover and mutation operations. Genetic algorithm selects individuals with high fitness from the total potential solution through iterative process, generates new individuals through crossover and mutation, and continuously optimizes until a satisfactory solution is found.

[0101] S243, using the trained genetic algorithm to calculate the installation layout and quantity of wall panels according to the wall surface data and wall panel size.

[0102] The principle of genetic algorithm to find the most suitable installation scheme is to simulate the genetic and natural selection mechanism in the process of biological evolution to gradually optimize the solution of the problem. Genetic algorithm simulates the evolution process of organisms in nature, and finds the optimal solution of the installation scheme through selection, crossover and mutation operations. The specific process includes: initialization of population: first, an initial population is randomly generated, each individual (or called chromosome) in the population represents a possible solution, a string of numbers or symbol sequences is used to form these individuals fitness evaluation: define a fitness function to evaluate the performance of each individual. The fitness function is designed according to the objective function of the problem, which is used to measure the adaptability of individuals to the problem, and individuals with high fitness have higher survival and reproduction opportunities. Selection operation: select a part of individuals from the population as the parents of the next generation according to the fitness function. There are many ways to select, such as roulette selection, tournament selection, tournament selection, etc. Crossover operation: the selected parent individuals are crossed to generate new offspring individuals. The crossover methods include single-point crossover, multi-point crossover, uniform crossover, etc. Mutation operation: in the offspring individuals generated after crossover, some genes are mutated with a certain probability to introduce new genetic information. Mutation generally uses random methods such as single-point mutation, multi-point mutation, bit mutation, etc. New population generation: new population is generated through selection, crossover and mutation operations, and the above process is repeated until the termination condition is met, such as reaching the maximum number of iterations or the fitness of the solution reaching the preset threshold.

[0103] Example 2

[0104] An image recognition-based wall panel defect intelligent detection system, comprising:

[0105] An image acquisition module for acquiring wall surface images;

[0106] A model optimization module for optimizing a wall panel installation management model;

[0107] An image processing module for inputting wall surface images into the wall panel installation management model, and the wall panel installation management model outputs the wall panel installation result.

[0108] Example 3

[0109] An electronic device comprises a chip, a processor and a memory, the memory is used for storing computer program code, the computer program code comprises computer instructions, under the condition that the chip executes the computer instructions, the electronic device executes an intelligent management method for wallboard installation.

[0110] Reference Figure 3 The electronic device 2 comprises a processor 21, a memory 22, an input device 23 and an output device 24. The processor 21, the memory 22, the input device 23 and the output device 24 are coupled through a connector, which comprises various interfaces, transmission lines or buses and the like, and the embodiments of the present application do not make any limitation. It should be understood that in various embodiments of the present application, coupling means mutual connection in a specific way, including direct connection or indirect connection through other devices, for example, various interfaces, transmission lines, buses and the like.

[0111] The processor 21 can be one or more graphics processing units (GPUs), and in the case that the processor 21 is a GPU, the GPU can be a single-core GPU or a multi-core GPU. Alternatively, the processor 21 can be a processor group composed of multiple GPUs, and the multiple processors are coupled with each other through one or more buses. Alternatively, the processor can also be other types of processors and the like, and the embodiments of the present application do not make any limitation.

[0112] The memory 22 can be used for storing computer program instructions, and various computer program codes for executing the program codes of the present application scheme. Alternatively, the memory comprises but is not limited to a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM) or a compact disc read-only memory (CD-ROM), which is used for related instructions and data.

[0113] The input device 23 is used for inputting data and / or signals, and the output device 24 is used for outputting data and / or signals. The output device 24 and the input device 23 can be independent devices, or can be an integral device.

[0114] Embodiment 4

[0115] A computer readable storage medium, the computer readable storage medium storing a computer program, the computer program comprising program instructions, the program instructions causing a processor to execute a kind of intelligent management method for wallboard installation when executed by the processor of an electronic device.

[0116] The wall image is preprocessed, and the installation position in the preprocessed wall image is easier to be detected, which can help machine vision to better identify the wallboard installation position. The wallboard installation management model is optimized and improved, so that the model is more sensitive to the wall installation position. The feature extraction network is replaced to better extract features related to the wallboard installation position. The loss function is optimized to make the detection accuracy of the model higher.

[0117] The above is only a specific embodiment of the present application, which enables those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features sought to be applied herein.

Claims

1. A method for intelligent management of wall panel installation, characterized by, The method comprises the following steps: acquiring a wall surface image; optimizing a wallboard installation management model; inputting the wall surface image into the wallboard installation management model, and outputting a wallboard installation result from the wallboard installation management model; wherein, after acquiring the wall surface image, the method further comprises pre-processing the wall surface image, specifically: graying the wall surface image; performing denoising processing on the grayed wall surface image; performing enhancement processing on the denoised wall surface image; performing size standardization on the enhanced wall surface image; wherein, the optimization of the wallboard installation management model comprises: replacing the architecture of the wallboard installation management model with ResNet50 feature extraction; adding RPN and Rol pooling detection layers to the wallboard installation management model; optimizing the loss function of the wallboard installation management model; adding a genetic algorithm to the wallboard installation management model; wherein, the optimization of the loss function of the wallboard installation management model comprises: In the formula, p i is the probability of the i-th anchor frame being predicted as the target, is the true value of the i-th anchor frame, which is 1 when it is a positive sample, otherwise 0, L1 is a classification loss function, L2 is a regression loss function, L t is the total loss function, N1 is the number of training anchor frames, N2 is the number of anchor frames that need to be regressed, α is a balance coefficient, smooth() is a smoothing function, t i is the offset parameter of the predicted frame and the corresponding anchor frame, is the offset parameter of the real target frame and the corresponding anchor frame; wherein, the addition of the genetic algorithm to the wallboard installation management model comprises: inputting wall surface data and wallboard size as a training set into the genetic algorithm; training the genetic algorithm; using the trained genetic algorithm to calculate the installation layout and number of wallboards according to the wall surface data and wallboard size.

2. The intelligent management method for wallboard installation of claim 1, wherein, after the denoising processing on the grayed wall surface image, the method further comprises judging whether the denoising effect meets the standard, specifically: the denoising effect is represented as: where f is the original image, g is the denoised image, u f is the mean of the original image, u g is the mean of the denoised image, σ (f，g) is the covariance of the original image and the denoised image, σ f is the variance of the original image, σ g is the variance of the denoised image, A1 and A2 are constants.

3. The intelligent management method for wallboard installation of claim 1, wherein, the replacement of the architecture of the wallboard installation management model with ResNet50 feature extraction comprises: using ResNet50 to perform feature extraction on the wall surface image, represented as: wherein w ij is a weight of the i-th row and j-th column of the convolution kernel, x ij is an element of the i-th row and j-th column of the input feature map, and b is a bias term, and the size of the convolution kernel is m x n.

4. An image recognition-based wallboard defect intelligent detection system, characterized in that, applying the method of any one of claims 1-3, comprising: an image acquisition module for acquiring a wall surface image; a model optimization module for optimizing a wallboard installation management model; an image processing module for inputting the wall surface image into the wallboard installation management model, and outputting a wallboard installation result from the wallboard installation management model.

5. An electronic device, comprising: The electronic device comprises a chip, a processor and a memory, the memory is used to store computer program code, the computer program code comprises computer instructions, and the electronic device executes the intelligent management method for wallboard installation according to any one of claims 1-3 when the chip executes the computer instructions. The computer readable storage medium stores a computer program, the computer program comprises program instructions, and the program instructions make the processor execute the intelligent management method for wallboard installation according to any one of claims 1-3 when the processor executes the program instructions.

6. A computer readable storage medium characterized by, ​

Images