Deep learning-based traffic jam detection system and detection method
Technical Field
The invention relates to a traffic jam detection system and a detection method, in particular to a deep learning-based traffic jam detection system for a surveillance video and a detection method applying the system, and belongs to the technical field of artificial intelligence.
Background
With the increasing of the vehicle retention rate and the developing of the automobile industry in China, the traffic pressure of various cities is rapidly increased at present, and the phenomenon of traffic jam frequently occurs in parts of regions, so that great challenges are brought to smooth operation of traffic in the cities, and various energy and environmental problems related to the traffic are aggravated. Therefore, the source and the propagation mechanism of the traffic jam are clear, the traffic jam is timely found and defibered, the smooth operation of urban traffic is guaranteed, and the method also gradually becomes the key research point in the industry.
In traditional traffic flow research, three basic characteristic parameters of traffic flow, namely flow, speed and density, are considered to exist, and the relationship among the three parameters is close. Through research for many years, the industry personnel find that the accuracy of the result is not high when the traffic state is judged by any single traffic flow parameter, so that the comprehensive research on the parameters becomes the mainstream of research, and a pairwise relation model among the three parameters is called as a relation model of traffic flow basic parameters.
At present, the research on the traffic jam detection and discrimination technology at home and abroad can be generally divided into the following categories according to the traffic parameter extraction technical method: the method comprises the steps of traffic parameter extraction based on a loop coil, traffic parameter extraction based on traffic video and traffic parameter extraction based on vehicle GPS data. However, in any of the above methods, the following problems may be more or less present in the application process. Firstly, the installation of specific equipment can damage a smooth road surface, the construction difficulty is high, and traffic jam is more easily caused in the construction process; secondly, a certain false detection rate exists no matter which method is adopted; thirdly, the research cost is high, and the scheme implementation difficulty and the management difficulty are high.
In recent years, with continuous development and maturation of artificial intelligence technology, related algorithms of artificial intelligence are continuously improved, chip performance is continuously improved, various manufacturers and enterprises start to research artificial intelligence algorithms, and under the background of the times, if the artificial intelligence technology can be applied to traffic jam detection, great technical breakthrough is certainly obtained.
In summary, how to provide a traffic jam detection system based on deep learning and a detection method using the system based on the prior art fills up a blank part in the current research field, and the problem that technical staff in the field need to solve is also a great need.
Disclosure of Invention
In view of the foregoing defects in the prior art, an object of the present invention is to provide a deep learning-based traffic congestion detection system for surveillance videos and a detection method using the system, which are described in detail below.
A deep learning based traffic congestion detection system comprising:
the video image acquisition module is used for acquiring moving vehicle information in the traffic monitoring video;
the video image recognition processing module is used for completing the tracking of moving vehicles in the traffic monitoring video, the acquisition of traffic characteristic parameters and the establishment of a traffic jam model;
the road congestion judging module is used for calculating a road congestion index and judging the congestion level of the road;
the video image acquisition module is connected with a plurality of traffic monitoring cameras in a road environment and acquires traffic monitoring videos by means of the traffic monitoring cameras;
the video image recognition processing module comprises a video processing server which is used for analyzing and processing the acquired traffic monitoring video so as to extract related traffic characteristic parameters;
the road congestion judging module comprises a calculating unit used for calculating a road congestion index and a judging unit used for judging the congestion level of the road according to the calculated road congestion index.
A deep learning-based traffic jam detection method uses the deep learning-based traffic jam detection system, and comprises the following steps:
s1, acquiring moving vehicle information in the traffic monitoring video by a moving target detection method;
s2, tracking the moving vehicle according to the information of the moving vehicle acquired in the S1;
s3, acquiring traffic characteristic parameters based on S2 and establishing a deep learning-based traffic congestion model;
and S4, calculating the road congestion index by using the traffic characteristic parameters in the S3 and judging the congestion level of the road.
Preferably, S1 specifically includes the following steps:
s11, finishing capturing video images from the traffic monitoring video;
s12, completing moving object detection from the captured video image;
and S13, identifying the vehicle model according to the detection result, returning to S11 again to capture the video image again if the moving object detected in S12 does not belong to the vehicle, and jumping to S2 if the moving object detected in S12 belongs to the vehicle.
Preferably, S2 specifically includes the following steps:
s21, detecting moving vehicles of moving targets of the video images;
s22, carrying out geometric correction on each moving vehicle;
s23, carrying out normalization processing on the moving vehicles after geometric correction to make the geometric positions of the moving vehicles uniform;
and S24, tracking the moving vehicle.
Preferably, S3 specifically includes the following steps:
s31, determining a vehicle speed V, which is an average speed of vehicles traveling on the road, based on the result in S2;
s32, determining a vehicle density D, which is a ratio of the number of vehicles traveling on the road to the road length, based on the result in S2;
s33, building a traffic jam model by utilizing the vehicle speed V and the vehicle density D, wherein the traffic jam model comprises an input layer, a plurality of hidden layers and an output layer, and the number of the hidden layers is multiple.
Preferably, the calculation formula of the vehicle speed V in S31 is:
where Vi denotes the traveling speed of the i-th vehicle, and n denotes the number of vehicles traveling on the road in the selected video image.
Preferably, the calculation formula of the vehicle density D in S32 is:
where m represents the number of lanes in the selected video image, L represents the length of the road in the selected video image, and n represents the number of vehicles traveling on the road in the selected video image.
Preferably, S4 specifically includes the following steps:
s41, calculating the traffic jam index delta according to the result in S3, wherein the calculation formula is as follows,
wherein V represents the vehicle speed, D represents the vehicle density, and P is a parameter and takes a value of 0.1;
and S42, judging the congestion level of the road according to the road congestion index delta obtained in S41, wherein the larger the traffic congestion index delta is, the more unblocked the road is, and the smaller the road congestion index delta is, the more congested the road is.
The advantages of the invention are mainly embodied in the following aspects:
the traffic jam detection system based on deep learning and the detection method applying the system can realize comprehensive, real-time and accurate perception of road traffic conditions, thereby effectively solving a plurality of problems existing in the traditional mode identification technology, meeting the increasing intelligent identification requirement of roads, and laying a solid foundation for improving the road management efficiency and ensuring the smooth operation of urban traffic.
Compared with the prior art, the method can realize automatic detection of the road traffic jam, not only obviously reduces the construction amount and the management cost, but also improves the accuracy of the traffic jam detection, and meanwhile, the method also has a plurality of functions of automatic alarm of traffic incidents, important position flow detection, easy blockage point detection and the like, and has excellent use effect.
In addition, the invention also provides reference for other related problems in the same field, can be expanded and extended on the basis of the reference, is applied to other technical schemes related to traffic jam detection in the same field, and has very wide application prospect.
The following detailed description of the embodiments of the present invention is provided in connection with the accompanying drawings for the purpose of facilitating understanding and understanding of the technical solutions of the present invention.
Drawings
FIG. 1 is a schematic structural diagram of a deep learning model;
FIG. 2 is a schematic flow diagram of the detection method of the present invention;
FIG. 3 is a schematic diagram of the architecture of the detection system of the present invention;
FIG. 4 is a diagram illustrating an embodiment of the present invention.
Detailed Description
The invention provides a traffic jam detection system based on deep learning and aiming at a monitoring video and a detection method applying the system, which can extract traffic characteristic parameters such as vehicle speed, lane number, vehicle density and the like from road traffic video monitoring, establish a road traffic jam evaluation index system, establish a traffic jam detection model based on deep learning according to the traffic jam evaluation index system, and finally automatically detect the road traffic jam according to the detection model. The details are as follows.
As shown in fig. 3, a deep learning-based traffic congestion detection system includes:
the video image acquisition module is used for acquiring moving vehicle information in the traffic monitoring video;
the video image recognition processing module is used for completing the tracking of moving vehicles in the traffic monitoring video, the acquisition of traffic characteristic parameters and the establishment of a traffic jam model;
the road congestion judging module is used for calculating a road congestion index and judging the congestion level of the road;
the video image acquisition module is connected with a plurality of traffic monitoring cameras in a road environment and acquires traffic monitoring videos by means of the traffic monitoring cameras;
the video image recognition processing module comprises a video processing server which is used for analyzing and processing the acquired traffic monitoring video so as to extract related traffic characteristic parameters;
the road congestion judging module comprises a calculating unit used for calculating a road congestion index and a judging unit used for judging the congestion level of the road according to the calculated road congestion index.
In the application process of the system, a video image acquisition module firstly acquires a traffic monitoring video through a large number of traffic monitoring cameras which are installed on a road; the video image recognition processing module analyzes and processes the acquired traffic monitoring video, and then extracts related traffic characteristic parameters such as vehicle speed, lane number and vehicle density; and finally, calculating a road congestion index by using the traffic congestion model through a road congestion judging module so as to judge the road congestion condition.
As shown in fig. 2, a deep learning-based traffic congestion detection method using the deep learning-based traffic congestion detection system includes the following steps:
and S1, acquiring the moving vehicle information in the traffic monitoring video by a moving target detection method.
And S2, tracking the moving vehicle according to the moving vehicle information acquired in the S1.
And S3, acquiring the traffic characteristic parameters based on the S2 and establishing a deep learning-based traffic congestion model.
And S4, calculating the road congestion index by using the traffic characteristic parameters in the S3 and judging the congestion level of the road.
Further, S1 specifically includes the following steps:
and S11, finishing capturing video images from the traffic monitoring video.
And S12, completing the detection of the moving object from the captured video image.
And S13, identifying the vehicle model according to the detection result, returning to S11 again to capture the video image again if the moving object detected in S12 does not belong to the vehicle, and jumping to S2 if the moving object detected in S12 belongs to the vehicle.
Further, S2 specifically includes the following steps:
and S21, detecting the moving vehicle of the moving object of the video image.
And S22, performing geometric correction on each moving vehicle.
And S23, carrying out normalization processing on the moving vehicles subjected to the geometric correction to make the geometric positions of the moving vehicles uniform.
And S24, tracking the moving vehicle.
Further, S3 specifically includes the following steps:
s31, determining a vehicle speed V, which is an average speed of vehicles traveling on the road, based on the result in S2; the calculation formula of the vehicle speed V in S31 is as follows,
where Vi denotes the traveling speed of the i-th vehicle, and n denotes the number of vehicles traveling on the road in the selected video image.
S32, determining a vehicle density D, which is a ratio of the number of vehicles traveling on the road to the road length, based on the result in S2; the calculation formula of the vehicle density D is as follows,
where m represents the number of lanes in the selected video image, L represents the length of the road in the selected video image, and n represents the number of vehicles traveling on the road in the selected video image.
S33, building a traffic jam model by utilizing the vehicle speed V and the vehicle density D, wherein the traffic jam model comprises an input layer, a plurality of hidden layers and an output layer, and the number of the hidden layers is multiple. The technical scheme of the invention is established on the basis of the deep learning model shown in figure 1, and more useful characteristics are learned by constructing a plurality of hidden layer models and massive training data, so that the accuracy of classification or prediction is finally improved.
Further, S4 specifically includes the following steps:
s41, calculating the traffic jam index delta according to the result in S3, wherein the traffic jam index delta refers to the degree of road jam, the more the vehicle speed is, the less the vehicle density is, the more the road is unblocked, and based on the above principle, the calculation formula of the traffic jam index delta is as follows,
wherein V represents the vehicle speed, D represents the vehicle density, and P is a parameter and takes a value of 0.1;
and S42, judging the congestion level of the road according to the road congestion index delta obtained in S41, wherein the larger the traffic congestion index delta is, the more unblocked the road is, and the smaller the road congestion index delta is, the more congested the road is.
In the technical solution of the present invention, the criterion of the road congestion degree is shown in table 1.
TABLE 1 Congestion level vs. Congestion index Range
The method adopts a traffic parameter extraction technology based on a traffic video and vehicle feature recognition and tracking based on a deep learning technology, so as to judge the vehicle speed and the vehicle density of the road and analyze the congestion condition of the road according to the established traffic congestion model. More useful features are learned by constructing a model with multiple hidden layers and massive training data, so that the accuracy of classification or prediction is finally improved. According to the invention, the automatic detection of the road traffic jam can be carried out, so that the construction amount is reduced, the management cost is reduced, and the accuracy of the traffic jam detection is improved.
In order to further verify the technical scheme of the invention, a section of a three-ring-line sister-in-law-village interchange portal frame east section which is easy to be congested is specially selected for 50 minutes at 16 pm in 21-week-day-year-9-2018, a test is selected, and according to the algorithm model training requirements, a plurality of groups of videos in different time periods are collected for analysis, and the result is shown in fig. 4 and table 2.
TABLE 2 Congestion analysis
The result shows that the detection effect of the algorithm is good. Through statistical analysis, the recognition rate of vehicle detection is about 94%, the false detection rate is low, and the result of the congestion model reaction basically accords with the actual road traffic condition, so that the method provided by the invention has a good detection effect.
The traffic jam detection system based on deep learning and the detection method applying the system can realize comprehensive, real-time and accurate perception of road traffic conditions, and have great breakthrough in recognition accuracy, recognition efficiency and recognition performance, so that various problems existing in the traditional mode recognition technology are effectively overcome, the increasing intelligent recognition requirements of roads are met, and a solid foundation is laid for improving the road management efficiency and guaranteeing the smooth operation of urban traffic.
Compared with the prior art, the method can realize automatic detection of the road traffic jam, not only remarkably reduces the construction amount and the management cost, but also greatly improves the accuracy of the traffic jam detection, and meanwhile, the method also has multiple functions of key position flow detection, easy blockage point detection and the like, and has excellent use effect.
In addition, the invention also provides reference for other related problems in the same field, can be expanded and extended on the basis of the reference, is applied to other technical schemes related to traffic jam detection in the same field, and has very wide application prospect.
In conclusion, the invention can be widely applied to road management scenes such as automatic alarm of traffic incidents, detection of flow of key positions, detection of easy blockage points and the like, and has very wide application prospect and application range.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein, and any reference signs in the claims are not intended to be construed as limiting the claim concerned.
Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.