CN112488026A - Lane damage detection method based on video analysis - Google Patents
Lane damage detection method based on video analysis Download PDFInfo
- Publication number
- CN112488026A CN112488026A CN202011452476.0A CN202011452476A CN112488026A CN 112488026 A CN112488026 A CN 112488026A CN 202011452476 A CN202011452476 A CN 202011452476A CN 112488026 A CN112488026 A CN 112488026A
- Authority
- CN
- China
- Prior art keywords
- sample
- roadway
- damaged
- video stream
- data set
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Abstract
A method for detecting the damage of a roadway based on video analysis comprises the steps of firstly obtaining sample pictures for forming a sample picture set, carrying out classification marking on the damaged road section area of the obtained sample pictures at the same time, and then carrying out model training on the classified and marked sample pictures to establish a sample picture Faster R-CNN model; the method comprises the steps of collecting a real-time video stream data set of the to-be-detected roadway, inputting the real-time video stream data set into a generated sample picture fast R-CNN model, judging the similarity between a target characteristic region and a training sample through the sample picture fast R-CNN model, and then determining whether the to-be-detected roadway has a damaged region or not, so that the problems that the traditional manual detection is time-consuming and labor-consuming, and the semi-automatic and automatic roadway detection algorithms are influenced by the environment and have the accuracy and the like are effectively solved, the accuracy and the real-time performance of the damage detection of the roadway are improved, the maintenance work efficiency is improved, and the maintenance.
Description
Technical Field
The invention relates to the technical field of pattern recognition, in particular to a method for detecting roadway damage based on video analysis.
Background
With the improvement of the traffic system in China, the number of motor vehicles is increasing day by day, and the safety of the driving lanes is closely related to urban traffic. The damage problems such as cracks and the like easily occur on the road surface of the roadway under the external influences of heavy pressure of vehicles, wind, rain and the like in the natural world. When the pavement is damaged by cracks and the like, the pavement needs to be maintained in time; otherwise, along with the increase of the damaged area of the road surface, the later maintenance cost is increased, and serious potential safety hazards are brought to the passing of vehicles. Therefore, how to detect whether the damaged area exists in the roadway becomes an urgent problem to be solved.
The traditional method for detecting the damage of the roadway mainly adopts manual detection, however, with the continuous increase of road networks and traffic flows, the method which takes time and expenses is not suitable, and the manual detection has the influence of decision subjectivity. In recent years, some semi-automatic and automatic roadway detection algorithms appear, which can eliminate man-made subjective influence, but are influenced by illumination conditions, roadway lines and other external environments, and the algorithms do not obtain good effect on accuracy and practicability, and the detection of the damage of the roadway is still a challenging task.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a method for detecting a damaged roadway based on video analysis, so as to solve the problems in the background art.
The technical problem solved by the invention is realized by adopting the following technical scheme:
a method for detecting the damage of a roadway based on video analysis comprises the steps of firstly obtaining sample pictures for forming a sample picture set, carrying out classification marking on the damaged road section area of the obtained sample pictures at the same time, and then carrying out model training on the classified and marked sample pictures to establish a sample picture Faster R-CNN model; collecting a real-time video stream data set of a to-be-detected roadway, inputting the real-time video stream data set into a generated sample picture Faster R-CNN model, judging the similarity between a target characteristic region and a training sample through the sample picture Faster R-CNN model, and determining whether the to-be-detected roadway has a damaged region; the method comprises the following specific steps:
1) collecting public damaged pictures of the roadway, or collecting pictures of at least one damaged road section of the actual roadway to form a sample picture set, classifying and marking damaged areas of sample pictures in the sample picture set, and marking corresponding sample damaged areas and damaged types in all the sample pictures;
2) respectively carrying out model training on the sample pictures of different damage areas and damage types obtained in the step 1) to establish a sample picture fast R-CNN model, training the model by adopting an optimized fast R-CNN method, wherein a feature network of the optimized fast R-CNN model uses a VGG-16 network structure processed by a compression algorithm, and the compression method is realized by combining channel pruning and low-order factorization;
3) after a camera collects a real-time video stream, the real-time video stream is transmitted to a server through a switch for video decoding, wherein the video stream is obtained by adopting FFmpeg, then the video stream is processed by CUDA decoding, and finally a decoded video stream data set is stored at a specified position;
4) storing the video stream data set in the step 3) into a sample picture fast R-CNN model generated by inputting the video stream data set at the specified position, wherein the model can determine the target damage type of a target characteristic area, and determining whether the to-be-detected roadway has a damaged area by judging the similarity between the target characteristic area and a training sample;
5) extracting at least one target characteristic area in a video stream data set by a sample picture fast R-CNN model, then calculating the characteristic similarity between the target characteristic area and a training sample, and determining whether a damaged area exists on the road surface of the to-be-detected roadway according to the extracted target characteristic area;
6) and when the similarity of at least one target characteristic area is greater than a set judgment threshold value, judging that the road surface of the roadway to be detected has a damaged area, using a red frame to circle the damaged area in the video data set, and sending out an alarm signal for warning.
In the invention, a sample training module, a real-time video module, a target extraction module and an alarm module are arranged in the sample picture Faster R-CNN model.
In the invention, a judgment threshold value is set in the sample picture Faster R-CNN model and is used for judging whether a damaged area exists in the extracted video stream data set.
Has the advantages that: according to the invention, the damage of the lane is detected through video analysis, so that the problems of time and labor waste, influence of environment on the accuracy of semi-automatic and automatic lane detection algorithms and the like in the traditional manual detection of the damage of the lane are effectively solved, the accuracy and the real-time performance of the detection of the damage of the lane are improved, the maintenance work efficiency is further improved, and the maintenance cost is saved.
Drawings
FIG. 1 is a flow chart illustrating a preferred embodiment of the present invention.
Detailed Description
In order to make the technical means, the creation characteristics, the achievement purposes and the effects of the invention easy to understand, the invention is further explained below by combining the specific drawings.
A method for detecting roadway damage based on video analysis comprises the following specific steps:
1) collecting public damaged pictures of the roadway, or collecting pictures of at least one damaged road section of the actual roadway to form a sample picture set, classifying and marking damaged areas of sample pictures in the sample picture set, and marking corresponding sample damaged areas and damaged types in all the sample pictures;
2) performing model training on the sample pictures of different damage areas and damage types acquired in the step 1) through a sample training module to establish a sample picture fast R-CNN model, training the model by adopting an optimized fast R-CNN method, wherein a feature network of the optimized fast R-CNN model uses a VGG-16 network structure processed by a compression algorithm, and the compression method is realized by combining channel pruning and low-order factor decomposition;
3) after a camera collects a real-time video stream, the real-time video stream is transmitted to a server through a switch for video decoding, wherein the video stream is obtained by adopting FFmpeg, then the video stream is processed by CUDA decoding, and finally a decoded video stream data set is stored at a specified position;
4) storing the step 3) into a sample picture Faster R-CNN model generated by inputting a video stream data set at a specified position through a real-time video module, wherein the model can determine the target damage type of a target characteristic area, and determining whether a damaged area exists in the to-be-detected roadway by judging the similarity between the target characteristic area and a training sample;
5) extracting at least one target characteristic area in a video stream data set by a target extraction module in a sample picture fast R-CNN model, then calculating the characteristic similarity between the target characteristic area and a training sample, and determining whether a damaged area exists on the road surface of the to-be-tested roadway according to the extracted target characteristic area;
6) and when the similarity of at least one target characteristic area is greater than a set judgment threshold value, judging that the road surface of the roadway to be detected has a damaged area, using a red frame to circle the damaged area in the video data set, and sending out an alarm signal for warning.
In this embodiment, the sample picture Faster R-CNN model is provided with a sample training module, a real-time video module, a target extraction module, and an alarm module.
In this embodiment, the sample picture Faster R-CNN model is provided with a determination threshold for determining whether a damaged area exists in the extracted video stream data set.
Claims (8)
1. A method for detecting the damage of a roadway based on video analysis is characterized in that sample pictures used for forming a sample picture set are obtained, at least one damaged road section is formed in each sample picture, meanwhile, classification marking is carried out on the damaged road section area of the obtained sample pictures, and then model training is carried out on the sample pictures after classification marking is finished so as to establish a fast R-CNN model of the sample pictures; the method comprises the steps of collecting a real-time video stream data set of a to-be-detected roadway, inputting the real-time video stream data set into a generated sample picture fast R-CNN model, judging the similarity between a target characteristic area and a training sample through the sample picture fast R-CNN model, and determining whether the to-be-detected roadway has a damaged area.
2. The method as claimed in claim 1, wherein the sample picture fast R-CNN model is provided with a sample training module, a real-time video module, and a target extraction module, and the sample picture fast R-CNN model is provided with a threshold for determining whether a damaged area exists in the extracted video stream data set.
3. The method for detecting the damage of the roadway based on the video analysis as claimed in any one of claims 1 to 2, characterized by comprising the following steps:
1) obtaining sample pictures for forming a sample picture set, classifying and marking damaged areas of the sample pictures, and marking corresponding sample damaged areas and damaged types in the sample pictures;
2) performing model training on the sample pictures of different damaged areas and damaged types obtained in the step 1) through a sample training module to establish a sample picture Faster R-CNN model;
3) collecting real-time video streams of a to-be-detected roadway, converting the video streams into a video stream data set, and storing the video stream data set at a specified position;
4) storing the step 3) into a sample picture Faster R-CNN model generated by inputting the video stream data set at the specified position through a real-time video module;
5) a target extraction module in a sample picture fast R-CNN model extracts at least one target characteristic area in a video stream data set, then calculates the characteristic similarity between the target characteristic area and a training sample, and determines whether a damaged area exists on the road surface of the to-be-tested roadway according to the extracted target characteristic area;
6) and when at least one target characteristic region exists, judging that the road surface of the to-be-detected roadway has a damaged region when the similarity of the target characteristic regions is greater than a set judgment threshold value.
4. The method as claimed in claim 3, wherein the step 1) is performed by collecting public damaged roadway images or collecting at least one damaged roadway section image of the actual roadway to form a sample image set.
5. The method for detecting roadway breakage based on video analysis as claimed in claim 3, wherein in step 2), model training is performed on the sample pictures of different breakage areas and breakage types obtained in step 1) by using an optimized Faster R-CNN method training model.
6. The method of claim 5, wherein the optimized fast R-CNN model feature network adopts a VGG-16 network structure processed by a compression algorithm.
7. The method according to claim 3, wherein in step 3), the real-time video stream of the roadway to be tested is collected by the camera and then transmitted to the server through the switch for video decoding, so as to convert the video stream into a video stream data set, and finally the video stream data set is stored at the specified position.
8. The method for detecting the damage of the roadway based on the video analysis as claimed in claim 2, wherein an alarm module is further disposed in the sample picture Faster R-CNN model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011452476.0A CN112488026A (en) | 2020-12-12 | 2020-12-12 | Lane damage detection method based on video analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011452476.0A CN112488026A (en) | 2020-12-12 | 2020-12-12 | Lane damage detection method based on video analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112488026A true CN112488026A (en) | 2021-03-12 |
Family
ID=74941550
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011452476.0A Pending CN112488026A (en) | 2020-12-12 | 2020-12-12 | Lane damage detection method based on video analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112488026A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113343891A (en) * | 2021-06-24 | 2021-09-03 | 深圳市起点人工智能科技有限公司 | Detection device and detection method for child kicking quilt |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9129355B1 (en) * | 2014-10-09 | 2015-09-08 | State Farm Mutual Automobile Insurance Company | Method and system for assessing damage to infrastructure |
CN107424150A (en) * | 2017-07-27 | 2017-12-01 | 济南浪潮高新科技投资发展有限公司 | A kind of road damage testing method and device based on convolutional neural networks |
CN108765404A (en) * | 2018-05-31 | 2018-11-06 | 南京行者易智能交通科技有限公司 | A kind of road damage testing method and device based on deep learning image classification |
-
2020
- 2020-12-12 CN CN202011452476.0A patent/CN112488026A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9129355B1 (en) * | 2014-10-09 | 2015-09-08 | State Farm Mutual Automobile Insurance Company | Method and system for assessing damage to infrastructure |
CN107424150A (en) * | 2017-07-27 | 2017-12-01 | 济南浪潮高新科技投资发展有限公司 | A kind of road damage testing method and device based on convolutional neural networks |
CN108765404A (en) * | 2018-05-31 | 2018-11-06 | 南京行者易智能交通科技有限公司 | A kind of road damage testing method and device based on deep learning image classification |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113343891A (en) * | 2021-06-24 | 2021-09-03 | 深圳市起点人工智能科技有限公司 | Detection device and detection method for child kicking quilt |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109977812B (en) | Vehicle-mounted video target detection method based on deep learning | |
Omer et al. | An automatic image recognition system for winter road surface condition classification | |
CN106874863B (en) | Vehicle illegal parking and reverse running detection method based on deep convolutional neural network | |
CN106845424B (en) | Pavement remnant detection method based on deep convolutional network | |
CN106373430B (en) | Intersection traffic early warning method based on computer vision | |
CN108765404A (en) | A kind of road damage testing method and device based on deep learning image classification | |
CN109255350B (en) | New energy license plate detection method based on video monitoring | |
CN113780312A (en) | Highway road surface condition detecting system | |
CN103034870B (en) | The boats and ships method for quickly identifying of feature based | |
CN104240239A (en) | Method for detecting local road segment hazy weather based on road image | |
Munawar | Image and video processing for defect detection in key infrastructure | |
CN110929676A (en) | Deep learning-based real-time detection method for illegal turning around | |
Jan et al. | A convolutional neural network based deep learning technique for identifying road attributes | |
CN111767874A (en) | Pavement disease detection method based on deep learning | |
CN112488026A (en) | Lane damage detection method based on video analysis | |
CN104268548A (en) | Accumulated snow detection method based on road images | |
CN116631187B (en) | Intelligent acquisition and analysis system for case on-site investigation information | |
CN111768373A (en) | Hierarchical pavement marking damage detection method based on deep learning | |
Danilescu et al. | Road anomalies detection using basic morphological algorithms | |
CN113191398A (en) | Improved Faster R-CNN-based pavement damage identification method | |
CN112802348A (en) | Traffic flow counting method based on mixed Gaussian model | |
CN112686956A (en) | Method for detecting tilt fault of urban road signal lamp post | |
CN113850111A (en) | Road condition identification method and system based on semantic segmentation and neural network technology | |
CN113850112A (en) | Road condition identification method and system based on twin neural network | |
CN113591814B (en) | Road cleanliness detection method and detection system based on dynamic vision |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210312 |