CN115908427A - Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM - Google Patents
Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM Download PDFInfo
- Publication number
- CN115908427A CN115908427A CN202310167135.6A CN202310167135A CN115908427A CN 115908427 A CN115908427 A CN 115908427A CN 202310167135 A CN202310167135 A CN 202310167135A CN 115908427 A CN115908427 A CN 115908427A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- improved
- svm
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Image Analysis (AREA)
Abstract
A pavement damage maintenance cost prediction method based on semantic segmentation and SVM comprises the following steps: step S1: acquiring a pavement image data set, segmenting a pavement image by utilizing semantics, and constructing a pre-training set, a training set and a testing set for network model training and testing; step S2: constructing an improved U-Net model; and step S3: pre-training the improved U-Net model on the pre-training set and re-training the pre-trained improved U-Net model on the training set; and step S4: storing the trained improved U-Net model, testing the trained improved U-Net model on a test set, and outputting a result; step S5: training the SVM model, and storing the SVM model; step S6: and extracting the pixel-level precision of the road surface diseases in the road surface images of the test set according to the improved U-Net model, and predicting the maintenance cost. The method has the advantages of high intelligent degree, improvement on prediction efficiency and accuracy and the like.
Description
Technical Field
The invention mainly relates to the technical field of traffic pavement image processing, in particular to a pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM.
Background
The highway transportation is flexible and convenient, the continuous improvement of the national highway transportation system promotes the high-speed development of economy, and the demand of the highway transportation system on the road operation and maintenance is continuously increased. The pavement can generate diseases such as transverse cracks, longitudinal cracks, crazing and the like due to factors such as increased use time, climate influence, improper use and maintenance modes and the like, and if the pavement diseases are not maintained in time, the service performance and safety performance of highway transportation can be influenced. Therefore, the method is very important for improving the reliability of highway transportation by quickly and accurately extracting pavement diseases.
Particularly for smart cities, the application of intelligent computing technologies such as internet of things, cloud computing, big data and space geographic information integration in the fields of city planning, design, construction, management and operation and the like through technical means is advocated and strived to enable key infrastructure components and services of cities such as city management, education, medical treatment, real estate, transportation, public utilities, public safety and the like to be more interconnected, efficient and intelligent, so that better life and work services are provided for citizens, a more favorable commercial development environment is created for enterprises, and a more efficient operation and management mechanism is enabled for governments.
The method for manually extracting the pavement diseases is long in time consumption, and the result is easily influenced by subjective experience. The image segmentation method based on the traditional features is used for segmenting a traffic road image according to features such as threshold features, similarity features, region edge features and the like extracted by an algorithm to extract road surface diseases, however, the method usually needs prior information, and has general segmentation precision for the traffic road image with a complex background, so that the method has limitation in extracting the road surface diseases. The semantic segmentation method based on deep learning is used for extracting the characteristics of a traffic road image according to a deep learning network model and conducting category prediction on each pixel point in the image so as to segment the traffic road image and achieve extraction of pixel level precision of road surface diseases in the image.
In recent years, semantic segmentation methods based on deep learning have been developed rapidly, and Full Convolutional Networks (FCN) proposed by Long et al, segNet proposed by Badrinarayanan et al, and U-Net proposed by Ronneberger et al have achieved excellent performance.
For example, the chinese patent application "road surface disease detection model training method, apparatus and computer device" (CN 112966665 a), in the technical solution includes: acquiring a plurality of road surface images; detecting a disease image in the plurality of road surface images; the disease image is a road surface image with diseases; a plurality of disease images are provided; marking the disease image according to the disease, and generating a mask image of the disease image according to marking content; obtaining a training data set of the pavement disease detection model according to the disease image and the mask image of the disease image; training the pavement disease detection model based on the training data set to obtain a trained pavement disease detection model; the trained pavement disease detection model is used for detecting pavement diseases of a pavement image to be detected. By adopting the method, all pavement diseases in the pavement image to be detected can be accurately identified.
At present, the method for predicting the maintenance cost of the road defects in the traffic road image is less, and all the existing intelligent image identification modes have the problems of low efficiency, low identification precision and the like, so that real unmanned, intelligent and reliable identification cannot be realized. Therefore, a method is needed to be provided, in which an extraction result of pavement disease pixel-level accuracy is obtained by semantically segmenting a traffic pavement image, and the maintenance cost of pavement diseases is predicted according to the extraction result of the pavement disease pixel-level accuracy.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems in the prior art, the invention provides the road surface disease maintenance cost prediction method and system based on semantic segmentation and SVM, which have the advantages of simple principle, high intelligence degree and capability of improving the prediction efficiency and accuracy.
In order to solve the technical problem, the invention adopts the following technical scheme:
a pavement damage maintenance cost prediction method based on semantic segmentation and SVM comprises the following steps:
step S1: acquiring a traffic road image data set, segmenting a traffic road image by utilizing semantics, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing the image in the pre-training set;
step S2: constructing an improved U-Net model;
and step S3: pre-training the improved U-Net model on a pre-training set, storing the pre-trained model, and training the pre-trained improved U-Net model again on the training set;
and step S4: storing the trained improved U-Net model, testing the trained improved U-Net model on a test set, and outputting a test result;
step S5: training the SVM model by using the marking of the pixel-level precision of the image in the training set and the cost of maintaining the pavement diseases in the image in the training set, and storing the trained SVM model;
step S6: and extracting the pixel-level precision of the road surface diseases in the traffic road surface images of the test set according to the improved U-Net model, and predicting the maintenance cost of the road surface diseases by using the trained SVM model.
As a further improvement of the process of the invention: in the step S1, the traffic road image data set adopts a GAPs384 data set obtained by processing an asphalt road damage data set.
As a further improvement of the process of the invention: each pixel in the image of the GAPs384 dataset represents 1.2mm × 1.2mm, and the image is cropped to 528 pixels in width and 432 pixels in height.
As a further improvement of the process of the invention: in the step S2, an asymmetric convolution structure and a dense void convolution structure are arranged in an encoder of the U-Net of the improved U-Net model, the asymmetric convolution structure is used to improve the extraction capability of the U-Net to the image features, the dense void convolution structure is used to extract the image features under different receptive fields, and the image features are fused with the image features extracted by the asymmetric convolution.
As a further improvement of the process of the invention: the outputs of the asymmetric convolution structure convolution layer and the hole convolution structure convolution layer activate function processing using a batch normalization BN and a modified linear unit ReLU.
As a further improvement of the process of the invention: in the modified U-Net model, 3 × 1 and 1 × 3 asymmetric convolutions are used instead of 3 × 3 convolutions with the same number of convolution kernels in the U-Net encoder.
As a further improvement of the process of the invention: in the improved U-Net model, the output characteristic diagram of a 1 × 3 convolutional layer with 32 convolutional kernels is processed by using dense void convolutional layers and a maximum pooling layer, and the output characteristic diagram of the 3 × 3 void convolutional layer with 64 convolutional kernels and 6 expansion rate is fused with the characteristic diagram of the 1 × 3 convolutional layer with 32 convolutional kernels, which is subjected to maximum pooling layer down-sampling; the feature graph of the output of the 3 × 3 cavity convolutional layer with the number of the convolution kernels of 128 and the expansion rate of 6 is fused with the feature graph of the output of the 1 × 3 convolutional layer with the number of the convolution kernels of 64 after the maximum pooling layer down-sampling; the feature map of the output of the 3 × 3 hole convolutional layer with the number of convolutional kernels of 256 and the expansion rate of 6 is fused with the feature map of the output of the 1 × 3 convolutional layer with the number of convolutional kernels of 128, which is subjected to maximum pooling layer down-sampling.
As a further improvement of the process of the invention: in the step S3, when the improved U-Net model is pre-trained and trained, the loss functionLUsing cross entropy loss functionL CE And a noise-robust Dice loss functionL NR-Dice The calculation formula is as follows:
in the formula (I), the compound is shown in the specification,λthe setting is made to be 0.8,Nrepresenting the number of pixel points in the traffic surface image,
indicates the th or fourth in the image>
The value marked correspondingly by each pixel point is combined>
Indicating an improved U-Net model versus the ^ th or greater in the image>
And predicting the value of each pixel point after passing through a Softmax function.
As a further improvement of the process of the invention: in the step S5, the SVM model is a non-linear SVM model.
The invention further provides a pavement damage maintenance cost predicting system based on semantic segmentation and SVM, which comprises:
the image data acquisition unit is used for acquiring a traffic road surface image data set, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing images in the pre-training set and the testing set;
the training unit is used for constructing an improved U-Net model; pre-training the improved U-Net model on a pre-training set, storing the pre-trained model, and training the pre-trained improved U-Net model again on the training set;
the test unit is used for storing the trained improved U-Net model, testing the trained improved U-Net model on a test set and outputting a test result;
the SVM model generating unit is used for training the SVM model by using the marking of the pixel-level precision of the image of the training set and the cost of maintaining the road surface diseases in the image of the training set, and storing the trained SVM model;
and the prediction unit is used for predicting the maintenance cost of the road defects by using the trained SVM model according to the extraction result of the improved U-Net model on the pixel-level precision of the road defects in the traffic road images of the test set.
Compared with the prior art, the invention has the advantages that: the road surface disease maintenance cost prediction method and system based on semantic segmentation and SVM have the advantages of simple principle, high intelligent degree and capability of improving the prediction efficiency and accuracy. According to the method, the traffic road image is segmented by utilizing semantics, and the maintenance cost of the road diseases is predicted by utilizing a Support Vector Machine (SVM) model according to the extraction result of the improved U-Net model on the pixel level precision of the road diseases in the traffic road image. According to the improved U-Net model, an asymmetric convolution and a dense void convolution structure are added in an encoder of the U-Net, and the extraction capability of the U-Net on image features is improved by using the asymmetric convolution. The method uses a dense cavity convolution structure to extract image characteristics under different receptive fields, and fuses the image characteristics with the image characteristics extracted by asymmetric convolution; and the decoder for improving U-Net outputs an extraction result of the pixel-level precision of the road surface diseases in the traffic road surface image according to the image characteristics extracted by the encoder, and predicts the maintenance cost of the road surface diseases according to the extraction result of the pixel-level precision of the road surface diseases by using an SVM model.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention.
Fig. 2 is a schematic diagram of the implementation principle of the invention in a specific application example.
Detailed Description
The invention will be described in further detail below with reference to the drawings and specific examples.
According to the method and the system for predicting the maintenance cost of the road surface diseases based on semantic segmentation and SVM, a traffic road surface image is segmented by utilizing the semantic, and the maintenance cost of the road surface diseases is predicted by using a Support Vector Machine (SVM) model according to the extraction result of an improved U-Net model on the pixel level precision of the road surface diseases in the traffic road surface image. According to the improved U-Net model, an asymmetric convolution and a dense void convolution structure are added in an encoder of the U-Net, and the extraction capability of the U-Net on image features is improved by using the asymmetric convolution. The method uses a dense cavity convolution structure to extract image characteristics under different receptive fields, and fuses the image characteristics with the image characteristics extracted by asymmetric convolution; the decoder of the improved U-Net outputs an extraction result of the pixel-level precision of the road surface diseases in the traffic road surface image according to the image characteristics extracted by the encoder, and an SVM model is used for predicting the maintenance cost of the road surface diseases according to the extraction result of the pixel-level precision of the road surface diseases.
As shown in fig. 1, the method for predicting the maintenance cost of pavement diseases based on semantic segmentation and SVM of the present invention comprises the following steps:
step S1: acquiring a traffic road surface image data set, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing images in the pre-training set and the testing set;
step S2: constructing an improved U-Net model;
and step S3: pre-training the improved U-Net model on a pre-training set, storing the pre-trained model, and training the pre-trained improved U-Net model again on the training set;
and step S4: storing the trained improved U-Net model, testing the trained improved U-Net model on a test set, and outputting a test result;
step S5: training the SVM model by using the marking of the pixel-level precision of the image of the training set and the cost of maintaining the pavement diseases in the image of the training set, and storing the trained SVM model;
step S6: and extracting the pixel-level precision of the road surface diseases in the traffic road surface images of the test set according to the improved U-Net model, and predicting the maintenance cost of the road surface diseases by using the trained SVM model.
In a specific application example, in the step S1, the traffic road image dataset may adopt a GAPs384 dataset obtained by processing an asphalt road damage dataset and/or a dataset actually acquired by the present invention according to actual needs; the GAPs384 dataset (GAPs) comprises 509 road surface disease images and corresponding image labels, and each pixel point in the images represents 1.2mm multiplied by 1.2mm. For pre-training, training and testing of the improved U-Net model, the invention in this example selects 509 images from the GAPs384 dataset as the pre-training set, as the preferred embodiment.
Further, in order to optimize the down-sampling and up-sampling processes of the network model on the image, the image is preferably clipped to have a width of 528 pixels and a height of 432 pixels.
In a specific application example, in the step S2, the invention constructs a structure of an improved U-Net model, as shown in fig. 2. In the figure, the position of the upper end of the main shaft,Nwhich represents the number of convolution kernels, is,dthe expansion rate of the convolution of the hole is expressed,S p the outputs representing the moving step size of the pooling layer pooling window, 3 × 1, 1 × 3, 3 × 3 convolutional layers and hole convolutional layers are processed using Batch Normalization (BN) and modified Linear Unit (ReLU) activation functions.
Furthermore, in the improved U-Net model, an asymmetric convolution and a dense void convolution structure are added in an encoder of the U-Net, and the extraction capability of the U-Net on image features is improved by using the asymmetric convolution; further, the invention uses the dense cavity convolution structure to extract the image characteristics under different receptive fields, and the image characteristics extracted by the asymmetric convolution are fused.
As a preferred embodiment, in this example, the innovation of the improved U-Net model is that it includes:
(a) The improved U-Net model replaces 3 x 3 convolutions with the same number of convolution kernels in the U-Net encoder with 3 x 1 and 1 x 3 asymmetric convolutions;
(b) The improved U-Net model processes the output characteristic diagram of the 1 x 3 convolutional layer with the number of convolutional kernels of 32 by using dense void convolutional layers and a maximum pooling layer, and the output characteristic diagram of the 3 x 3 void convolutional layer with the number of convolutional kernels of 64 and the expansion rate of 6 is fused with the characteristic diagram of the 1 x 3 convolutional layer with the number of convolutional kernels of 32, which is subjected to maximum pooling layer down-sampling; further, the output feature map of the 3 × 3 hole convolutional layer with the number of convolution kernels of 128 and the expansion rate of 6 is fused with the feature map of the 1 × 3 convolutional layer with the number of convolution kernels of 64 after maximum pooling layer down-sampling; the feature map of the output of the 3 × 3 hole convolutional layer with the number of convolutional kernels of 256 and the expansion rate of 6 is fused with the feature map of the output of the 1 × 3 convolutional layer with the number of convolutional kernels of 128, which is subjected to maximum pooling layer down-sampling.
In a specific application example, in the step S3, when the improved U-Net model is pre-trained and trained, adam is used by the optimizer, the learning rate is set to 0.001, the batch size is set to 5, and 100 rounds of training are performed on the pre-training set.
In a specific application example, in the step S3, when the improved U-Net model is pre-trained and trained, the loss function is usedLUsing cross entropy loss functionL CE And a noise-robust Dice loss functionL NR-Dice The calculation formula is as follows:
in the formula (I), the compound is shown in the specification,λthe setting is made to be 0.8,Nrepresents the number of pixel points in the traffic road surface image,
indicates the th or fourth in the image>
The value marked correspondingly by each pixel point is combined>
Representing an improved U-Net model vs. a ^ th or a greater than th in an image>
And predicting the value of each pixel point after passing through a Softmax function.
In a specific application example, in the step S5, a non-linear SVM model is adopted as the SVM model.
The invention further provides a pavement disease maintenance cost predicting system based on semantic segmentation and SVM, which comprises:
the image data acquisition unit is used for acquiring a traffic road surface image data set, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing images in the pre-training set and the testing set;
the training unit is used for constructing an improved U-Net model; pre-training the improved U-Net model on the pre-training set, storing the pre-trained model, and re-training the pre-trained improved U-Net model on the training set;
the test unit is used for storing the trained improved U-Net model, testing the trained improved U-Net model on a test set and outputting a test result;
the SVM model generating unit is used for training the SVM model by using the marking of the pixel-level precision of the image of the training set and the cost of maintaining the pavement diseases in the image of the training set, and storing the trained SVM model;
and the prediction unit is used for predicting the maintenance cost of the road defects by using the trained SVM model according to the extraction result of the improved U-Net model on the pixel-level precision of the road defects in the traffic road images of the test set.
As will be appreciated by one skilled in the art, the above-described embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (10)
1. A pavement disease maintenance cost prediction method based on semantic segmentation and SVM is characterized by comprising the following steps: the method comprises the following steps:
step S1: acquiring a traffic road image data set, segmenting a traffic road image by utilizing semantics, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing the image in the pre-training set and the training set;
step S2: constructing an improved U-Net model;
and step S3: pre-training the improved U-Net model on the pre-training set, storing the pre-trained model, and re-training the pre-trained improved U-Net model on the training set;
and step S4: storing the trained improved U-Net model, testing the trained improved U-Net model on a test set, and outputting a test result;
step S5: training the SVM model by using the marking of the pixel-level precision of the image in the training set and the cost of maintaining the pavement diseases in the image in the training set, and storing the trained SVM model;
step S6: and extracting the pixel-level precision of the road surface diseases in the traffic road surface images of the test set according to the improved U-Net model, and predicting the maintenance cost of the road surface diseases by using the trained SVM model.
2. The road surface disease maintenance cost prediction method based on semantic segmentation and SVM according to claim 1, characterized by comprising the steps of: in the step S1, the traffic road image data set adopts a GAPs384 data set obtained by processing an asphalt road damage data set.
3. The road surface damage repair cost prediction method based on semantic segmentation and SVM according to claim 2, characterized in that: each pixel in the image of the GAPs384 dataset represents 1.2mm x 1.2mm, and the image is cropped to 528 pixels wide and 432 pixels high.
4. The road surface damage maintenance cost prediction method based on semantic segmentation and SVM according to any one of claims 1-3, characterized by comprising the following steps: in the step S2, an asymmetric convolution structure and a dense void convolution structure are arranged in an encoder of the U-Net of the improved U-Net model, the asymmetric convolution structure is used to improve the extraction capability of the U-Net to the image features, the dense void convolution structure is used to extract the image features under different receptive fields, and the image features are fused with the image features extracted by the asymmetric convolution.
5. The road surface damage repair cost prediction method based on semantic segmentation and SVM according to claim 4, characterized in that: the outputs of the convolution layer of the asymmetric convolution structure and the hole convolution layer of the hole convolution structure activate function processing using batch normalization BN and modified linear unit ReLU.
6. The road surface disease maintenance cost prediction method based on semantic segmentation and SVM according to claim 5, characterized in that: in the modified U-Net model, 3 × 1 and 1 × 3 asymmetric convolutions are used instead of 3 × 3 convolutions with the same number of convolution kernels in the U-Net encoder.
7. The road surface damage repair cost prediction method based on semantic segmentation and SVM according to claim 5, characterized in that: in the improved U-Net model, the output characteristic diagram of a 1 × 3 convolutional layer with 32 convolutional kernels is processed by using dense void convolutional layers and a maximum pooling layer, and the output characteristic diagram of the 3 × 3 void convolutional layer with 64 convolutional kernels and 6 expansion rate is fused with the characteristic diagram of the 1 × 3 convolutional layer with 32 convolutional kernels, which is subjected to maximum pooling layer down-sampling; the feature graph of the output of the 3 × 3 cavity convolutional layer with the number of the convolution kernels of 128 and the expansion rate of 6 is fused with the feature graph of the output of the 1 × 3 convolutional layer with the number of the convolution kernels of 64 after the maximum pooling layer down-sampling; the output feature maps of 3 × 3 hole convolutional layers with 256 convolutional kernel numbers and 6 expansion ratios and the feature maps of the outputs of 1 × 3 convolutional layers with 128 convolutional kernel numbers, which were subjected to maximum pooling layer down-sampling, were fused.
8. The road surface disease maintenance cost prediction method based on semantic segmentation and SVM according to claim 5, characterized in that: in the step S3, when the improved U-Net model is pre-trained and trained, the loss functionLUsing cross entropy loss functionL CE And a noise-robust Dice loss functionL NR-Dice The calculation formula is as follows:
in the formula (I), the compound is shown in the specification,λthe setting is made to be 0.8,Nrepresents the number of pixel points in the traffic road surface image,
indicates the th or fourth in the image>
The value of each pixel point corresponding to the label is greater or less>
Indicating an improved U-Net model versus the ^ th or greater in the image>
And predicting the value of each pixel point after passing through the Softmax function.
9. The road surface disease maintenance cost prediction method based on semantic segmentation and SVM according to any one of claims 1-3, characterized by comprising the steps of: in the step S5, the SVM model is a non-linear SVM model.
10. A road surface disease maintenance cost predicting system based on semantic segmentation and SVM is characterized by comprising the following components:
the image data acquisition unit is used for acquiring a traffic road surface image data set, constructing a pre-training set, a training set and a testing set for network model training and testing, and preprocessing images in the pre-training set and the testing set;
the training unit is used for constructing an improved U-Net model; pre-training the improved U-Net model on a pre-training set, storing the pre-trained model, and training the pre-trained improved U-Net model again on the training set;
the test unit is used for storing the trained improved U-Net model, testing the trained improved U-Net model on a test set and outputting a test result;
the SVM model generating unit is used for training the SVM model by using the marking of the pixel-level precision of the image of the training set and the cost of maintaining the road surface diseases in the image of the training set, and storing the trained SVM model;
and the prediction unit is used for predicting the maintenance cost of the road defects by using the trained SVM model according to the extraction result of the improved U-Net model on the pixel-level precision of the road defects in the traffic road images of the test set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310167135.6A CN115908427B (en) | 2023-02-27 | 2023-02-27 | Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310167135.6A CN115908427B (en) | 2023-02-27 | 2023-02-27 | Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115908427A true CN115908427A (en) | 2023-04-04 |
| CN115908427B CN115908427B (en) | 2023-05-12 |
Family
ID=85737462
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310167135.6A Active CN115908427B (en) | 2023-02-27 | 2023-02-27 | Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908427B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130155061A1 (en) * | 2011-12-16 | 2013-06-20 | University Of Southern California | Autonomous pavement condition assessment |
| CN109801282A (en) * | 2019-01-24 | 2019-05-24 | 湖北大学 | Pavement behavior detection method, processing method, apparatus and system |
| CN110020686A (en) * | 2019-04-10 | 2019-07-16 | 浙江大学城市学院 | A kind of road surface method for detecting abnormality based on intelligent perception sensing data |
| CN112949783A (en) * | 2021-04-29 | 2021-06-11 | 南京信息工程大学滨江学院 | Road crack detection method based on improved U-Net neural network |
| CN115100207A (en) * | 2022-08-26 | 2022-09-23 | 北京恒新天创科技有限公司 | Detection system and detection method based on machine vision |
-
2023
- 2023-02-27 CN CN202310167135.6A patent/CN115908427B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130155061A1 (en) * | 2011-12-16 | 2013-06-20 | University Of Southern California | Autonomous pavement condition assessment |
| CN109801282A (en) * | 2019-01-24 | 2019-05-24 | 湖北大学 | Pavement behavior detection method, processing method, apparatus and system |
| CN110020686A (en) * | 2019-04-10 | 2019-07-16 | 浙江大学城市学院 | A kind of road surface method for detecting abnormality based on intelligent perception sensing data |
| CN112949783A (en) * | 2021-04-29 | 2021-06-11 | 南京信息工程大学滨江学院 | Road crack detection method based on improved U-Net neural network |
| CN115100207A (en) * | 2022-08-26 | 2022-09-23 | 北京恒新天创科技有限公司 | Detection system and detection method based on machine vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115908427B (en) | 2023-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110322495B (en) | Scene text segmentation method based on weak supervised deep learning | |
| CN110610166B (en) | Text region detection model training method and device, electronic equipment and storage medium | |
| CN109753913B (en) | Multi-mode video semantic segmentation method with high calculation efficiency | |
| Dornaika et al. | Building detection from orthophotos using a machine learning approach: An empirical study on image segmentation and descriptors | |
| CN110555433B (en) | Image processing method, device, electronic equipment and computer readable storage medium | |
| CN111104903B (en) | Depth perception traffic scene multi-target detection method and system | |
| CN112084901A (en) | GCAM-based high-resolution SAR image airport runway area automatic detection method and system | |
| CN110879960B (en) | Method and computing device for generating image data set for convolutional neural network learning | |
| CN112232237B (en) | Method, system, computer device and storage medium for monitoring vehicle flow | |
| CN103049763A (en) | Context-constraint-based target identification method | |
| CN110717886A (en) | Pavement pool detection method based on machine vision in complex environment | |
| CN112232351A (en) | License plate recognition system based on deep neural network | |
| Xu et al. | HA U-Net: Improved model for building extraction from high resolution remote sensing imagery | |
| CN116012815A (en) | Traffic element identification method, multi-task network model, training method and training device | |
| CN110222772B (en) | Medical image annotation recommendation method based on block-level active learning | |
| Lin et al. | Semi-supervised lane detection with deep hough transform | |
| CN113704276A (en) | Map updating method and device, electronic equipment and computer readable storage medium | |
| CN114943834B (en) | Full-field Jing Yuyi segmentation method based on prototype queue learning under few labeling samples | |
| CN116740495A (en) | Training method and defect detection method for defect detection model of road and bridge tunnel | |
| CN115908427B (en) | Pavement disease maintenance cost prediction method and system based on semantic segmentation and SVM | |
| CN115861324B (en) | Pavement damage maintenance cost prediction method and system | |
| CN114241470A (en) | Natural scene character detection method based on attention mechanism | |
| CN114202765A (en) | Image text recognition method and storage medium | |
| Dursa et al. | Developing traffic congestion detection model using deep learning approach: a case study of Addis Ababa city road | |
| Li et al. | Chinese License Plate Recognition Algorithm Based On UNet3+ |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |