Pavement surface skid resistance prediction method based on image recognition

Abstract

The invention provides a pavement surface skid resistance prediction method based on image recognition. The pavement surface anti-skid performance prediction method based on image recognition comprises the steps of a, image acquisition: acquiring image data of a pavement surface through high-definition camera equipment or unmanned aerial vehicle equipment, and b, image preprocessing: aiming at the acquired image data, adopting an image enhancement algorithm to obtain a prediction result of the pavement surface anti-skid performance. I, image enhancement: performing enhancement processing on the low-illumination image by using an adaptive histogram equalization algorithm to improve the image contrast, and ii, denoising processing: removing noise by using a Gaussian filtering method to ensure the definition of the image. According to the pavement surface anti-skid performance prediction method based on image recognition, a self-adaptive histogram equalization algorithm is used for enhancing a low-illumination image and improving the contrast ratio of the image, so that the visibility of details is effectively improved, and a Gaussian filtering denoising method is used for removing noise caused by environmental factors.

Description

Technical Field

[0001] The present invention relates to the technical field of road surface anti-skid performance prediction, and in particular to a road surface anti-skid performance prediction method based on image recognition. Background Art

[0002] The image recognition-based pavement surface anti-skid performance prediction method consists of three core modules: image acquisition, image processing and feature extraction, and anti-skid performance prediction. First, the image acquisition module acquires image data of the pavement surface using a high-definition camera or drone. Then, the image processing and feature extraction module uses image processing techniques to process the acquired images and extract key surface features such as cracks, wear, and potholes. Finally, the anti-skid performance prediction module uses machine learning or deep learning algorithms based on the extracted feature information to predict the pavement's anti-skid performance, providing a scientific basis for road maintenance decisions.

[0003] While this method theoretically holds promise, it also suffers from certain drawbacks. First, image quality significantly impacts prediction results. In low light, inclement weather, or underperforming equipment, images may be blurry or missing, leading to inaccurate feature extraction and, in turn, impacting the accuracy of skid resistance prediction. Second, image data processing and feature extraction require significant computing resources, particularly when processing large-scale road surface data. This can lead to issues such as low processing efficiency and high latency. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method for predicting the anti-skid performance of road surfaces based on image recognition, which solves the problem that images may be blurred or missing in low light, bad weather or poor equipment performance, resulting in inaccurate feature extraction; it requires a large amount of computing resources, especially when processing large-scale road data, and may face problems of low processing efficiency and high time delay.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for predicting the anti-skid performance of a road surface based on image recognition, comprising:

[0006] a. Image acquisition: Obtain image data of the road surface through high-definition cameras or drones;

[0007] b. Image preprocessing: Image enhancement algorithm is used for the collected image data:

[0008] i. Image enhancement: Use adaptive histogram equalization algorithm to enhance low-light images to improve image contrast;

[0009] ii. Denoising: Use Gaussian filtering to remove noise and ensure image clarity;

[0010] iii. Blur restoration: Uses convolutional neural networks to restore blurred areas, using deep learning algorithms to automatically repair blurred or missing areas in images;

[0011] c. Feature extraction: Perform feature extraction on the processed image and use a convolutional neural network model to extract key features in the image, such as road surface texture, cracks, potholes, and wear. The output of the feature extraction model is a feature vector F = [f1, f2, ..., f n ], where f i represents the i-th extracted feature, and n is the total number of features;

[0012] d. Anti-skid performance modeling: Based on the extracted image features, an anti-skid performance prediction model is constructed using the following model formula:

[0013] P slip =σ(W T F+b)

[0014] Among them, P slip Represents the predicted value of the road surface anti-skid performance, W is the weight vector of the model, F is the image feature vector, b is the bias term, and σ is the activation function, usually the Sigmoid function

[0015] e. Performance prediction: Based on the above anti-skid performance model, calculate and output the anti-skid performance prediction value P of the road surface slip , in order to assess whether there is a risk of slippery road surface;

[0016] f. Model optimization and accuracy improvement: The model parameters W and b are adjusted through the back propagation algorithm to improve the accuracy of anti-slip performance prediction. The gradient descent method is used to minimize the loss function L, which is:

[0017]

[0018] in, is the true anti-slip performance value, m is the number of samples, is the predicted anti-skid performance value;

[0019] g. Large-scale data processing and acceleration: Adopt a distributed computing framework, optimize the data processing process, use multi-level data parallel processing and efficient feature extraction algorithms to improve the efficiency of road image data processing and reduce time delays.

[0020] Preferably, the image enhancement algorithm in step b is an adaptive histogram equalization method based on limited contrast to further improve the image quality under low light conditions.

[0021] Preferably, the convolutional neural network model in step c adopts a residual network structure, which can effectively extract road features of different scales.

[0022] Preferably, the anti-skid performance prediction model in step d uses a support vector machine as a classifier, combined with deep learning feature output, to perform final classification of the anti-skid performance.

[0023] Preferably, the model optimization uses an adaptive learning rate adjustment strategy to accelerate the convergence process and prevent overfitting.

[0024] Preferably, the large-scale data processing process realizes rapid analysis of a large number of road surface images through batch processing of image data and compression coding of image features.

[0025] Preferably, the loss function in step f is a weighted square error loss function, wherein the error of each sample is assigned different weights according to its importance in the evaluation of the anti-skid performance of the road surface.

[0026] Preferably, the method further includes performing segmented predictions based on different road conditions through a multi-stage prediction strategy to improve the prediction accuracy in different scenarios. In the process of predicting the anti-skid performance of the road surface, the method adopts time series data fusion technology, combines historical data and real-time image data, and improves the robustness of the model.

[0027] The present invention provides a method for predicting the anti-skid performance of a road surface based on image recognition. It has the following beneficial effects:

[0028] This image recognition-based method for predicting the skid resistance of road surfaces uses an adaptive histogram equalization algorithm to enhance low-light images, improving image contrast and effectively enhancing the visibility of details. It also employs a Gaussian filter denoising method to remove noise caused by environmental factors. It also uses a convolutional neural network to repair blurred areas and a deep learning algorithm to automatically repair any blurred or missing areas in the image, ensuring high-quality image clarity and providing high-quality image data for subsequent feature extraction.

[0029] In the process of image data processing and feature extraction, the present invention automatically extracts key features of the pavement surface through a convolutional neural network model, including pavement texture, cracks, potholes, and wear, etc. These features are closely related to the anti-skid performance of the pavement. Unlike traditional methods, the feature extraction process of the present invention is not only accurate but also efficient, avoiding the complexity and time delay problems of manual feature extraction in traditional methods. In addition, after feature extraction, the proposed anti-skid performance prediction model uses a deep learning algorithm to model the anti-skid performance of the pavement surface and outputs accurate prediction results. By optimizing the use of computing resources, especially for the processing of large-scale pavement data, the present invention can effectively reduce computing resource consumption and significantly improve prediction speed and accuracy, and is suitable for large-scale data processing in actual pavement management and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting the anti-skid performance of a road surface based on image recognition, including: a. Image acquisition: obtaining image data of the road surface through a high-definition camera device or a drone device.

[0033] b. Image preprocessing: Image enhancement algorithm is used for the collected image data:

[0034] i. Image enhancement: Use the adaptive histogram equalization algorithm to enhance low-light images to improve image contrast.

[0035] ii. Denoising: Use Gaussian filtering to remove noise to ensure image clarity.

[0036] iii. Blur restoration: A convolutional neural network is used to restore blurred areas, and a deep learning algorithm is used to automatically repair blurred or missing areas in the image. The image enhancement algorithm is based on a contrast-limited adaptive histogram equalization method to further improve image quality under low-light conditions.

[0037] c. Feature extraction: Perform feature extraction on the processed image and use a convolutional neural network model to extract key features in the image, such as road surface texture, cracks, potholes, and wear. The output of the feature extraction model is a feature vector F = [f1, f2, ..., f n ], where f i represents the i-th extracted feature, n is the total number of features, and the convolutional neural network model adopts a residual network structure, which can effectively extract road features of different scales.

[0038] d. Anti-skid performance modeling: Based on the extracted image features, an anti-skid performance prediction model is constructed using the following model formula:

[0039] P slip =σ(W T F+b)

[0040] Among them, P slip Represents the predicted value of the road surface anti-skid performance, W is the weight vector of the model, F is the image feature vector, b is the bias term, and σ is the activation function, usually the Sigmoid function The anti-skid performance prediction model uses support vector machine as a classifier and combines deep learning feature output to perform the final classification of anti-skid performance.

[0041] e. Performance prediction: Based on the above anti-skid performance model, calculate and output the anti-skid performance prediction value P of the road surface slip , in order to assess whether there is a risk of slippery road surface.

[0042] f. Model optimization and accuracy improvement: The model parameters W and b are adjusted through the back propagation algorithm to improve the accuracy of anti-slip performance prediction. The gradient descent method is used to minimize the loss function L, which is:

[0043]

[0044] in, is the true anti-slip performance value, m is the number of samples, To predict the anti-skid performance value, the loss function is the weighted square error loss function, in which the error of each sample is given different weights according to its importance in the anti-skid performance evaluation of the pavement. The large-scale data processing process realizes the rapid analysis of a large number of pavement images through batch processing of image data and compression encoding of image features.

[0045] g. Large-scale data processing and acceleration: A distributed computing framework is adopted to optimize the data processing process. Multi-level data parallel processing and efficient feature extraction algorithms are used to improve the efficiency of road image data processing and reduce time delays. The method further includes a multi-stage prediction strategy to perform segmented predictions based on different road conditions to improve the prediction accuracy in different scenarios. In the process of predicting the anti-skid performance of the road surface, the method adopts time series data fusion technology, combines historical data and real-time image data, and improves the robustness of the model. The large-scale data processing process realizes rapid analysis of a large number of road images through batch processing of image data and compression encoding of image features.

[0046] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the anti-skid performance of a road surface based on image recognition, characterized in that: include: a. Image acquisition: Obtain image data of the road surface through high-definition camera equipment or drone equipment; b. Image preprocessing: Image enhancement algorithm is used for the collected image data: i. Image enhancement: Use adaptive histogram equalization algorithm to enhance low-light images to improve image contrast; ii. Denoising: Use Gaussian filtering method to remove noise to ensure image clarity; iii. Blur repair: Use convolutional neural networks to repair blurred areas, and use deep learning algorithms to automatically repair blurred or missing areas in images; c. Feature extraction: Perform feature extraction on the processed image and use a convolutional neural network model to extract key features in the image, such as road surface texture, cracks, potholes and wear. The output of the feature extraction model is a feature vector F = [ f1,f2,…,f n] , where f i represents the i-th extracted feature, and n is the total number of features; d. Anti-skid performance modeling: Based on the extracted image features, an anti-skid performance prediction model is constructed using the following model formula: P slip =σ ( W T F+b Among them, P slip represents the predicted value of the road surface anti-skid performance, W is the weight vector of the model, F is the image feature vector, b is the bias term, and σ is the activation function, usually the Sigmoid function e. Performance prediction: According to the above anti-skid performance model, calculate and output the anti-skid performance prediction value P of the road surface slip , in order to assess whether there is a risk of slippery road surface; f. Model optimization and accuracy improvement: The model parameters, W and b, are adjusted through the back propagation algorithm to improve the accuracy of anti-slip performance prediction. The gradient descent method is used to minimize the loss function L, which is: in, is the true anti-slip performance value, m is the number of samples, is the predicted anti-skid performance value; g. Large-scale data processing and acceleration: Adopt a distributed computing framework, optimize the data processing process, use multi-level data parallel processing and efficient feature extraction algorithms to improve the efficiency of road image data processing and reduce time delays.

2. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The image enhancement algorithm in step b is an adaptive histogram equalization method based on limited contrast to further improve the image quality under low light conditions.

3. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The convolutional neural network model in step c adopts a residual network structure, which can effectively extract road features of different scales.

4. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The anti-skid performance prediction model described in step d uses a support vector machine as a classifier, combined with deep learning feature output, to perform a final classification of the anti-skid performance.

5. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The model optimization uses an adaptive learning rate adjustment strategy to accelerate the convergence process and prevent overfitting.

6. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The large-scale data processing process realizes rapid analysis of a large number of road surface images through batch processing of image data and compression coding of image features.

7. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The loss function in step f is a weighted square error loss function, in which the error of each sample is given different weights according to its importance in the evaluation of the anti-skid performance of the road surface.

8. The method for predicting the anti-skid performance of a road surface based on image recognition according to claim 1, characterized in that: The method further includes performing segmented prediction according to different road conditions through a multi-stage prediction strategy to improve the prediction accuracy in different scenarios. In the process of predicting the anti-skid performance of the road surface, the method uses time series data fusion technology to combine historical data and real-time image data to improve the robustness of the model.

Images