Self-adaptive intelligent sensing data processing method and system for intelligent construction site

Abstract

The invention relates to the technical field of data processing, in particular to a self-adaptive intelligent sensing data processing method and system for an intelligent construction site. The method comprises the steps of collecting document image data; preprocessing the acquired image data; performing feature extraction on the image data by using a convolutional neural network (CNN); analyzing the weights of the extracted different features by adopting an entropy weight method; and optimizing network training by reconstructing a loss function. Through the Gaussian filtering preprocessing technology, the random noise interference in the image is significantly reduced, the feature extraction precision of the CNN model under the conditions of common paper stain and damage, wrinkles and the like in the construction site environment is enhanced, and when the image quality is poor, the high recognition accuracy and robustness can still be maintained.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an adaptive intelligent perception data processing method and system for a smart construction site. Background Art

[0002] With the development of technology, the construction of smart construction sites has gradually become an important measure to improve the quality of construction site management, and it has become particularly important to enter various information materials of workers. Since the mid-twentieth century, image recognition technology, as an important branch of artificial intelligence, has evolved from traditional images to deep learning algorithms and has been widely used in all walks of life. With the rise of deep learning technology, especially the development of convolutional neural networks (CNN), the accuracy and efficiency of image recognition have been greatly promoted.

[0003] However, the existing technology of using CNN models to recognize images still has many problems in use on construction sites. Due to the harsh environment on site and the impact of workers' work, paper materials inevitably have problems such as stains and wrinkles, and the CNN model cannot effectively identify the original text and the content of the stained area, resulting in material entry errors and increasing the burden of construction site management. At the same time, CNN model training requires a lot of computing resources, and most construction sites have limited hardware equipment and usually cannot provide high-performance hosts, which is also a huge challenge for accurate material recognition and efficient entry. Therefore, improving the accuracy and clarity of image recognition in harsh conditions on construction sites is of great significance to the construction of smart construction sites. Summary of the invention

[0004] In order to solve the above-mentioned problems, the present invention provides an adaptive intelligent perception data processing method and system for smart construction sites. It aims to improve the existing image recognition technology, and proposes an adaptive intelligent perception data processing method for smart construction site applications in combination with the actual situation on the construction site. Gaussian filtering is used to pre-process the scanned document image to reduce image noise and improve the feature extraction quality of the CNN model. At the same time, it can simplify the image details, reduce the number of parameters and calculation amount of the CNN model to a certain extent, and improve the recognition speed. The entropy weight method is introduced to adjust the weights of different features, which can provide a more reasonable weight distribution for the loss function of the CNN model, making the training process more efficient and improving the recognition accuracy of the model, so as to accurately and efficiently enter personnel information, improve work efficiency, and optimize user experience.

[0005] In the first aspect, the present invention provides an adaptive intelligent perception data processing method for a smart construction site, which adopts the following technical solutions: An adaptive intelligent perception data processing method for a smart construction site, comprising: Collecting document image data; Preprocessing the acquired image data; Use convolutional neural network CNN to extract features from image data; The entropy weight method is used to analyze the weights of different extracted features; Optimize network training by reconstructing the loss function.

[0006] Furthermore, the collecting of document image data includes collecting multiple images of data, including tables, texts and photos of various document materials required for entry to the construction site.

[0007] Furthermore, the preprocessing of the acquired image data includes performing Gaussian filtering on the acquired document image data, wherein the Gaussian filtering process is expressed as: In the formula, is the coordinate within the kernel relative to the center point, is the standard deviation of the Gaussian distribution, is the original image dataset, is the filtered image dataset, is the Gaussian kernel radius, is the location of the Gaussian kernel The value of For the original image at position The pixel value of .

[0008] Furthermore, the feature extraction of image data using a convolutional neural network (CNN) includes sequentially performing convolution operations and pooling operations on the image data, wherein a data feature set is extracted through the convolution operation, and a dimension of the data feature set is reduced through the pooling operation.

[0009] Furthermore, the feature extraction of the image data using the convolutional neural network CNN also includes converting the data feature set after the convolution operation and the pooling operation into a one-dimensional data feature vector set, and outputting a predicted value, wherein the output predicted value is calculated using a Softmax function, and the one-dimensional data feature vector set It is expressed as: In the formula, ( ) is the activation function, is the weight matrix, is the bias vector.

[0010] Furthermore, the entropy weight method is used to analyze the weights of different extracted features, including standardizing the extracted data features so that they have the same dimension and distribution, and constructing a decision matrix for the standardized features, wherein each column of the decision matrix represents a feature, each row represents a sample, and the feature weight of each data feature is calculated.

[0011] Furthermore, the optimization of network training by reconstructing the loss function includes taking the cross entropy loss function and the feature weight result as the basis for network training evaluation, and the cross entropy function expression is: In the formula, is the true value, is the predicted value.

[0012] In the second aspect, an adaptive intelligent perception data processing system for smart construction sites includes: The data acquisition module is configured to collect document image data; A preprocessing module is configured to preprocess the acquired image data; The feature extraction module is configured to extract features from the image data using a convolutional neural network (CNN); The entropy weight module is configured to analyze the weights of different extracted features using an entropy weight method; The training module is configured to optimize the network training by reconstructing the loss function.

[0013] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for example, a method for processing adaptive intelligent perception data for a smart construction site.

[0014] In a fourth aspect, the present invention provides a terminal device including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to implement the adaptive intelligent perception data processing method for smart construction sites.

[0015] In summary, the present invention has the following beneficial technical effects: 1. Through Gaussian filtering preprocessing technology, the random noise interference in the image is significantly reduced, and the feature extraction accuracy of the CNN model is enhanced in common situations such as paper stains and wrinkles in construction site environments. When the image quality is poor, it can still maintain a high degree of recognition accuracy and robustness.

[0016] 2. The preprocessing step of Gaussian filtering not only denoises the image, but also appropriately simplifies the image details, effectively reducing the parameter scale and computing resources required for the CNN model, ensuring that high-speed and accurate recognition performance can be maintained in hardware-constrained construction sites, and realizing rapid information entry and processing.

[0017] 3. The entropy weight method is used to dynamically adjust the weight ratio of each feature in the CNN model, so that the model can focus on key features for training. Even when the image is partially damaged, the model can still accurately identify the core information, thereby significantly improving the stability of the recognition method. 4. The automated and intelligent identification mechanism has liberated manpower to a great extent, improved the efficiency and reliability of data collection, thereby promoting real-time monitoring and fine management of engineering projects, and optimizing the overall operation mode and user experience of smart construction sites.

[0018] 5. The rapid response and high-precision information entry functions ensure the smooth flow of project information, avoid the risks caused by data lags or errors, and lay a solid foundation for the efficient operation of smart construction sites. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of an adaptive intelligent perception data processing method for a smart construction site according to Example 1 of the present invention. DETAILED DESCRIPTION

[0020] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0021] Example 1 Reference Figure 1 , an adaptive intelligent perception data processing method for a smart construction site in this embodiment includes: Collecting document image data; Preprocessing the acquired image data; Use convolutional neural network CNN to extract features from image data; The entropy weight method is used to analyze the weights of different extracted features; Optimize network training by reconstructing the loss function.

[0022] Specifically: Step 1: Collect multiple images, including forms, text, photos and other documents required for entry to the construction site. The paper has varying degrees of stains and wrinkles.

[0023] Step 2: Perform Gaussian filtering on the image data to remove random noise and make the image smoother. Use a Gaussian convolution kernel to scan the pixels of the image. The Gaussian convolution kernel is expressed as: In the formula, is the coordinate within the kernel relative to the center point, is the standard deviation of the Gaussian distribution.

[0024] The Gaussian filtering process can be expressed as: In the formula, is the original image dataset, is the filtered image dataset, is the Gaussian kernel radius, is the location of the Gaussian kernel The value of For the original image at position The pixel value of .

[0025] Step 3: Use convolutional neural network (CNN) to extract features from image data.

[0026] Step 3-1: First, perform convolution operation on the collected image data. The input image data set is , the convolution kernel is , the convolution operation formula is: In the formula, represents the convolution operation, is a pixel point of the output feature map, is the convolution kernel weight. ( ) is the activation function, and its expression is: Step 3-2: To reduce the dimension of the feature map while retaining important information, perform a pooling operation on the data feature set obtained after the convolution operation. The pooling operation formula is: In the formula, is the center of the pooling window.

[0027] Step 3-3: Set the data feature set after convolution and pooling operations Transform into a one-dimensional data feature vector set .

[0028] In the formula, ( ) is the activation function, is the weight matrix, is the bias vector.

[0029] The model output prediction value is Function calculation: Step 4: Use the entropy weight method to determine the weights of different features. First, the extracted data features are standardized to have the same dimension and distribution for easy comparison.

[0030] Constructing a decision matrix using standardized features , each column represents a feature, and each row represents a sample. Let the total number of samples be , for each feature , calculate its entropy : In the formula, For the The sample in The ratio of features, .

[0031] Let the total number of features be , calculate the coefficient of variation for each feature and weight : Step 5: Combine the cross entropy loss function with the feature weights as the basis for network training evaluation. The cross entropy function expression is: In the formula, is the true value, is the predicted value.

[0032] Combined with feature weight analysis, the proportion of important features is increased and the proportion of unimportant features is reduced. The improved loss function is: Step 6: Use the loss function combined with the entropy weight method to quantify the gap between the predicted value and the true value, and guide the neural network to optimize the prediction results.

[0033] Example 2 This embodiment provides an adaptive intelligent perception data processing system for a smart construction site, including: The data acquisition module is configured to collect document image data; A preprocessing module is configured to preprocess the acquired image data; The feature extraction module is configured to extract features from the image data using a convolutional neural network (CNN); The entropy weight module is configured to analyze the weights of different extracted features using an entropy weight method; The training module is configured to optimize the network training by reconstructing the loss function.

[0034] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for the adaptive intelligent perception data processing method for a smart construction site.

[0035] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to implement the adaptive intelligent perception data processing method for smart construction sites.

[0036] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An adaptive intelligent perception data processing method for smart construction sites, characterized in that: include: Collect document image data; Preprocessing the acquired image data; Use convolutional neural network CNN to extract features from image data; The entropy weight method is used to analyze the weights of different extracted features; Optimize network training by reconstructing the loss function.

2. The method for processing adaptive intelligent perception data for smart construction sites according to claim 1, characterized in that: The collecting of document image data includes collecting multiple image data, including tables, texts and photos of various document materials required for entry into the construction site.

3. The method for processing adaptive intelligent perception data for smart construction sites according to claim 2 is characterized in that: The preprocessing of the acquired image data includes performing Gaussian filtering on the acquired document image data, wherein the Gaussian filtering process is expressed as: In the formula, is the coordinate within the kernel relative to the center point, is the standard deviation of the Gaussian distribution, is the original image dataset, is the filtered image dataset, is the Gaussian kernel radius, is the location of the Gaussian kernel The value of For the original image at position The pixel value of .

4. The method for processing adaptive intelligent perception data for smart construction sites according to claim 3 is characterized in that: The feature extraction of image data using a convolutional neural network (CNN) includes sequentially performing convolution operations and pooling operations on the image data, wherein a data feature set is extracted through the convolution operation, and a dimension of the data feature set is reduced through the pooling operation.

5. The method for processing adaptive intelligent perception data for smart construction sites according to claim 4 is characterized in that: The method of using a convolutional neural network (CNN) to extract features from image data also includes converting the data feature set after the convolution operation and the pooling operation into a one-dimensional data feature vector set, and outputting a predicted value, wherein the output predicted value is calculated using a Softmax function, and the one-dimensional data feature vector set It is expressed as: In the formula, ( ) is the activation function, is the weight matrix, is the bias vector.

6. The method for processing adaptive intelligent perception data for smart construction sites according to claim 5 is characterized in that: The entropy weight method is used to analyze the weights of different extracted features, including standardizing the extracted data features so that they have the same dimension and distribution, and constructing a decision matrix for the standardized features, wherein each column of the decision matrix represents a feature, each row represents a sample, and the feature weight of each data feature is calculated.

7. The method for processing adaptive intelligent perception data for smart construction sites according to claim 6, characterized in that: The optimization of network training by reconstructing the loss function includes taking the cross entropy loss function and the feature weight result as the basis for network training evaluation. The cross entropy function expression is: In the formula, is the true value, is the predicted value.

8. An adaptive intelligent perception data processing system for smart construction sites, characterized in that: include: A data acquisition module is configured to collect document image data; A preprocessing module is configured to preprocess the acquired image data; The feature extraction module is configured to extract features from the image data using a convolutional neural network (CNN); The entropy weight module is configured to analyze the weights of different extracted features using an entropy weight method; The training module is configured to optimize the network training by reconstructing the loss function.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the method according to claim 1 .

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the method as claimed in claim 1 .

Images