CN114202572A - Vehicle load measuring and transportation route planning method and system based on machine vision - Google Patents
Vehicle load measuring and transportation route planning method and system based on machine vision Download PDFInfo
- Publication number
- CN114202572A CN114202572A CN202210145462.7A CN202210145462A CN114202572A CN 114202572 A CN114202572 A CN 114202572A CN 202210145462 A CN202210145462 A CN 202210145462A CN 114202572 A CN114202572 A CN 114202572A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- tire
- transportation
- load
- axle
- 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
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000008569 process Effects 0.000 claims description 29
- 230000004044 response Effects 0.000 claims description 29
- 238000012549 training Methods 0.000 claims description 28
- 238000004422 calculation algorithm Methods 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 21
- 238000013528 artificial neural network Methods 0.000 claims description 19
- 210000002569 neuron Anatomy 0.000 claims description 19
- 238000013461 design Methods 0.000 claims description 16
- 210000000349 chromosome Anatomy 0.000 claims description 13
- 230000003068 static effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000005516 engineering process Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 9
- 230000002068 genetic effect Effects 0.000 claims description 8
- 238000012216 screening Methods 0.000 claims description 8
- 238000001514 detection method Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 108090000623 proteins and genes Proteins 0.000 claims description 6
- 238000003708 edge detection Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 2
- 239000000126 substance Substances 0.000 claims description 2
- 238000005259 measurement Methods 0.000 abstract description 10
- 239000000284 extract Substances 0.000 abstract description 3
- 238000007726 management method Methods 0.000 description 10
- 238000012544 monitoring process Methods 0.000 description 10
- 230000035772 mutation Effects 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000005303 weighing Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011056 performance test Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005183 dynamical system Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005297 material degradation process Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
- G06Q10/0835—Relationships between shipper or supplier and carriers
- G06Q10/08355—Routing methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/052—Detecting movement of traffic to be counted or controlled with provision for determining speed or overspeed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30236—Traffic on road, railway or crossing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Strategic Management (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Geometry (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Game Theory and Decision Science (AREA)
- Primary Health Care (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a vehicle load measuring and transport route planning method and system based on machine vision, which extracts the actual deformation, the tire model and the tire pressure information of a tire corresponding to each axle from a tire image; inputting the actual deformation, the tire model, the tire pressure information and the position distribution data of each axle tire of the vehicle to be measured into a trained vehicle load identification model to obtain the axle weight data of each axle of the vehicle to be measured, and calculating the load data of the vehicle to be measured based on the axle weight data of each axle, so that the load measurement efficiency can be greatly improved; in addition, the invention screens the transportation routes of the vehicles by analyzing the passability of the transportation routes and calculating the bearing capacity of the bridge, so that the screened transportation routes can better ensure the safety of cargo transportation.
Description
Technical Field
The invention relates to the technical field of vehicle transportation management, in particular to a vehicle load measuring and transportation route planning method and system based on machine vision.
Background
Different roads have different load bearing capacities due to different construction standards, so that large cargos with the characteristics of overweight, overlong, superwide, undetachable and the like need to finish the transportation process through a road bridge section of a specific route. The transportation of the large cargos comprises the processes of large cargo transportation declaration, line selection, trafficability inspection and safety evaluation, examination and approval, transportation condition monitoring and the like, and relates to the problems of large cargo truck load identification, path planning selection, bridge bearing capacity checking, reinforcement treatment and the like. The bridge bearing capacity determines the trafficability of the route, and if the bridge bearing capacity on the trafficway cannot meet the design requirement, the load and the bridge are damaged slightly, and serious safety accidents can be caused seriously. Therefore, before transporting goods, the large transportation enterprises need to submit declaration transportation plans to relevant departments, and the inspection institutions and the approval institutions perform safety evaluation and inspection on the large transportation process. If the inspection result is that the route does not meet the passing requirement, the transportation enterprise needs to change the transportation route or the transportation mode, and then needs to submit and declare the transportation plan again for route evaluation. However, the transportation of large pieces involves multiple departments, and because the bridges on the route are built in different times, the used design theory, specification, material performance and later maintenance conditions have great differences, and the checking and calculating process and the approval process of the bearing capacity of the bridges take a long time.
In addition, the large transportation enterprises need to provide the information of the overall dimensions of the truck and the goods during the processes of axle weight, dead weight, load and cargo loading when reporting. The axle load information of the existing large transport vehicle is only obtained by weighing instruments such as static wagon balance and the like, the instruments are not portable, and the weighing efficiency of the large transport vehicle, namely a multi-axle multi-tire special vehicle, is low.
How to reasonably and quickly complete the identification of the weight of the large-sized cargo truck, route planning and review of the truck weight is a technical problem to be solved by the technical personnel in the field.
Disclosure of Invention
The invention provides a vehicle load measuring and transport route planning method and system based on machine vision, which are used for solving the technical problems of low weighing efficiency and low route planning efficiency of a large-piece freight car.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a vehicle load measuring method based on machine vision comprises the following steps:
collecting tire images of the left side and the right side of a vehicle to be tested, and acquiring the number of tires on each axle of the vehicle to be tested;
extracting the actual deformation, the tire model and the tire pressure information of the tire corresponding to each axle from the tire image;
inputting the actual deformation, the tire model, the tire pressure information and the number of tires on the axle of the vehicle to be tested into a trained vehicle load identification model to obtain the axle weight data of each axle of the vehicle to be tested, and calculating the load data of the vehicle to be tested based on the axle weight data of each axle.
Preferably, the vehicle load identification model is an artificial neural network model, the artificial neural network model includes a multilayer artificial neural network, each neuron in the layer is not connected, the neurons in the layer are fully connected, a weighted value of each neuron adopts random initialization, and a Sigmoid function is used as an activation function of each neuron.
Preferably, the training process of the vehicle load model is as follows:
constructing a training set of the vehicle load model, taking the training set as input of a training network, carrying out forward propagation, comparing a network output value with an expected output value, and calculating an error value; taking the error value as a learning basis, and reversely propagating the error layer by using a gradient descent method to realize the adjustment of the neuron parameter of each layer;
the learning rate is adaptively adjusted during the training of the vehicle load model, wherein the adjustment rule is as follows:
wherein the content of the first and second substances,
is the sum of the squared errors for the k-th iteration,
the learning rate for the kth iteration.
Preferably, the step of extracting the actual deformation amount, the tire model number and the tire pressure information of the tire corresponding to each axle from the tire image includes the steps of:
identifying the model, the rim diameter and the tire pressure information of each tire in the tire image based on a character identification algorithm; identifying each axle, the corresponding tire of each axle, the tire contour and the rim contour of each tire from the tire image based on a Mask R-CNN identification algorithm, and extracting the tire contour and the rim contour of each tire from the tire image by adopting an edge detection measuring method;
dividing the tire contour into an upper tire contour and a lower tire contour by taking a horizontal straight line passing through the circle center of the rim contour as a dividing line; calculating the difference between the pixel points of the upper tire contour and the lower tire contour as the tire deformation pixel value; and calculating a scale factor according to the identified rim diameter and the number of the rim diameter pixel points in the image, and calculating the actual deformation of the tire according to the scale factor and the tire deformation pixel value.
A vehicle transportation route planning method based on machine vision comprises the following steps:
calculating load data of the vehicle to be planned by adopting the method, and determining an origin place and a target place of the vehicle to be planned;
determining a transportation network topology between an origin and a target, acquiring the stability requirement of a vehicle to be planned and the contour parameters of the vehicle to be planned to determine the feasible parameter range of roads and bridges of a feasible transportation line;
acquiring parameters of roads and bridges of each transportation road section in a transportation network topology, comparing the parameters of the roads and bridges of each transportation road section with a feasible parameter range, and selecting the transportation road section with the parameters of the roads and bridges in the feasible parameter range as a feasible transportation road section;
acquiring vehicle standard load of a feasible transportation road section, comparing the vehicle standard load of the feasible transportation road section with the load data of the vehicle to be planned, and selecting the feasible transportation road section with the vehicle standard load larger than the load data of the vehicle to be planned as all safe transportation road sections of the vehicle to be planned;
and constructing a safe transportation network topology according to all safe transportation road sections of the vehicle to be planned, and screening out an optimal transportation line from the safe transportation network topology.
Preferably, the parameters of the road and the bridge include: the road sign board height, the minimum flat curve radius of the route, the maximum longitudinal slope and the maximum transverse slope are combined;
further comprising the steps of:
identifying the wheelbase of the vehicle to be planned by adopting a machine vision technology, and constructing a load model of the vehicle to be planned by combining the axle weight of each axle of the vehicle to be planned;
when the standard load of the vehicle is smaller than the load data of the vehicle to be planned, calling pre-constructed structural data of the bridge from a finite element database, constructing a finite element calculation model of the bridge according to the structural data, and loading the load model of the vehicle to be planned on the finite element calculation model of the bridge according to the actual running position of the vehicle to be planned on the bridge, so as to obtain static and dynamic response values of the bridge under the load action of the vehicle to be planned;
comparing the static and dynamic response values with the design specification response of the bridge, if the static and dynamic response values are smaller than the design specification response values, judging that the feasible transportation road section corresponding to the bridge is a safe transportation road section, and if the static and dynamic response values are larger than the design specification response values and not larger than 1.05 times of the design specification response values, judging that the feasible transportation road section after the bridge is reinforced is the safe transportation road section; and if the static and dynamic response values are larger than the 1.05 design specification response value, judging that the feasible transportation road section corresponding to the bridge is not a safe transportation road section.
Preferably, the method for identifying the wheelbase of the vehicle to be planned by using the machine vision technology comprises the following steps:
acquiring a vehicle image simultaneously containing two axle wheels, identifying two adjacent wheels from the vehicle image based on a Mask R-CNN identification algorithm, and detecting the rim profiles of the two adjacent wheels from the vehicle image by adopting an edge detection method; marking the circle center positions of the rims of the two adjacent wheels, and calculating a pixel point value between the circle center positions of the rims of the two adjacent wheels;
for any one wheel A of two adjacent wheels, identifying the rim diameter of the wheel A from a vehicle image based on a character identification technology, and calculating a rim diameter pixel point of the wheel A in the vehicle image; calculating a scale factor according to the identified rim diameter and the number of rim diameter pixel points in the image; and multiplying the pixel point value between the circle center positions of the wheel rims of the two adjacent wheels by the scaling factor to obtain the wheel base of the two adjacent wheels.
Preferably, the method for screening out the optimal transportation route from the safe transportation network topology specifically comprises the following steps:
constructing a target optimization function which takes a safe transportation line of a safe transportation network topology as an independent variable, and takes the minimum total distance of the safe transportation line and the maximum safety factor of the bridge;
encoding each transport node in the secure transport network topology; coding each safe transportation line between the origin and the target site based on the codes of each transportation node to obtain a coded safe transportation line set;
taking each encoded safe transportation line as a chromosome, taking the encoded transportation nodes as genes of the chromosome, and taking the target optimization function as a fitness function to construct a genetic algorithm to solve the optimal solution of the target optimization function; and taking the safe transportation line corresponding to the optimal solution as an optimal transportation line.
Preferably, the method further comprises the following steps:
constructing and training a wheel identification model taking YOLOv5 as a frame, acquiring a vehicle video of a vehicle in a transportation process through a camera, inputting the vehicle video into the wheel identification model to obtain vehicle videos of frames marked with wheel prediction frames, associating the prediction frames of the same wheel in the frames of the vehicle video by adopting a Deepsort algorithm, and allocating IDs to different wheels according to a sequence to realize the tracking of the vehicle;
extracting the position of a pixel point of the middle point of the lower rectangular part of the wheel prediction frame as the position of a contact point of the wheel and the bridge floor, and calculating the actual position of the vehicle according to the shooting angle and the focal length of a camera, the position of the camera and the position of the contact point;
recording the time stamp of each tire of the vehicle leaving the preset virtual detection area during the tracking process, and calculating the average speed of the vehicle by using the wheel base between two wheels and the time difference of the wheels leaving the detection area.
A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method when executing the computer program.
The invention has the following beneficial effects:
1. the vehicle load measuring and transportation route planning method and system based on machine vision extract the actual deformation, the tire model and the tire pressure information of the corresponding tire of each axle from the tire image; inputting the actual deformation, the tire model, the tire pressure information and the number of tires of each axle tire of the vehicle to be measured into a trained vehicle load identification model to obtain the axle weight data of each axle of the vehicle to be measured, and calculating the load data of the vehicle to be measured based on the axle weight data of each axle, so that the load measurement efficiency can be greatly improved; in addition, the invention screens the transportation routes of the vehicles by analyzing the passability of the transportation routes and calculating the bearing capacity of the bridge, so that the screened transportation routes can better ensure the safety of cargo transportation.
2. In the preferred scheme, the vehicle tire parameters are identified based on the machine vision technology, the axle load prediction model is established based on the artificial neural network, the identified vehicle tire parameters are input into the axle load prediction model, the axle load data of the vehicle are obtained, and the load data of the vehicle is calculated according to the axle load data, so that the vehicle load measurement automation can be realized while the vehicle load measurement precision is ensured, and the vehicle load measurement speed and convenience are greatly improved.
3. In the preferred scheme, the method encodes the nodes of the road network based on a genetic algorithm, represents all selectable routes by chromosomes, constructs an objective function by taking the shortest distance and traffic safety as targets, screens off partial filial generation individuals through selection, intersection and variation operations, and finally obtains a theoretical optimal path to realize path optimization; the method can improve the efficiency of path planning, improve the safety of vehicle transportation and reduce the cost of vehicle transportation.
4. In a preferable scheme, the tire is identified based on a machine vision technology, the actual position of the vehicle is accurately determined, and the real-time speed monitoring of the vehicle is realized.
In addition to the objects, features and advantages described above, other objects, features and advantages of the present invention are also provided. The present invention will be described in further detail below with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
fig. 1 is a flow chart of a method for measuring vehicle load and planning transportation route based on machine vision in the preferred embodiment of the invention;
FIG. 2 is a flow chart of a method for machine vision based vehicle load measurement in a preferred embodiment of the present invention;
FIG. 3 is a diagram of an artificial neural network hierarchy in a preferred embodiment of the present invention;
FIG. 4 is a schematic numbering diagram of the nodes of the routing network in the preferred embodiment of the present invention;
fig. 5 is a flow chart of a transportation route planning method based on machine vision in a preferred embodiment of the present invention.
Detailed Description
The embodiments of the invention will be described in detail below with reference to the drawings, but the invention can be implemented in many different ways as defined and covered by the claims.
The first embodiment is as follows:
the implementation discloses a vehicle load measuring method based on machine vision, which comprises the following steps:
collecting tire images of the left side and the right side of a vehicle to be detected, and acquiring the number of tires on one side of each axle;
extracting the actual deformation, the tire model and the tire pressure information of the tire corresponding to each axle from the tire image;
and inputting the actual deformation, the tire model, the tire pressure information and the number of the single-side tires on each axle of the vehicle to be detected into a trained vehicle load identification model to obtain the axle weight data of each axle of the vehicle to be detected, and calculating the load data of the vehicle to be detected based on the axle weight data of each axle.
In addition, in the present embodiment, as shown in fig. 5, a vehicle transportation route planning method based on machine vision is also disclosed, which includes the following steps:
calculating load data of the vehicle to be planned by adopting the method, and determining an origin place and a target place of the vehicle to be planned;
determining a transportation network topology between an origin and a target, acquiring the stability requirement of a vehicle to be planned and the contour parameters of the vehicle to be planned to determine the feasible parameter range of roads and bridges of a feasible transportation line;
acquiring parameters of roads and bridges of each transportation road section in a transportation network topology, comparing the parameters of the roads and bridges of each transportation road section with a feasible parameter range, and selecting the transportation road section with the parameters of the roads and bridges in the feasible parameter range as a feasible transportation road section;
acquiring vehicle standard load of a feasible transportation road section, comparing the vehicle standard load of the feasible transportation road section with the load data of the vehicle to be planned, and selecting the feasible transportation road section with the vehicle standard load larger than the load data of the vehicle to be planned as all safe transportation road sections of the vehicle to be planned;
and constructing a safe transportation network topology according to all safe transportation road sections of the vehicle to be planned, and screening out an optimal transportation line from the safe transportation network topology.
In addition, in the embodiment, a computer system is also disclosed, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of the method are implemented.
The vehicle load measuring and transportation route planning method and system based on machine vision extract the actual deformation, the tire model and the tire pressure information of the corresponding tire of each axle from the tire image; inputting the actual deformation, the tire model, the tire pressure information and the number of single-side tires on each axle of the vehicle to be measured into a trained vehicle load identification model which is input and trained to obtain the axle weight data of each axle of the vehicle to be measured, and calculating the load data of the vehicle to be measured based on the axle weight data of each axle, thereby greatly improving the efficiency of load measurement; in addition, the invention screens the transportation routes of the vehicles by analyzing the passability of the transportation routes and calculating the bearing capacity of the bridge, so that the screened transportation routes can better ensure the safety of cargo transportation.
Example two:
the second embodiment is the preferred embodiment of the first embodiment, and the difference from the first embodiment is that the specific steps of the vehicle load measurement and transportation route planning method based on the machine vision are refined, and the fusion application of the vehicle load measurement and transportation route planning method based on the machine vision is introduced:
in the present embodiment, as shown in fig. 1, a method for measuring vehicle load and planning transportation route based on machine vision is disclosed, which includes the following steps:
s1, determining a transportation network topology between an origin place and a target place of the vehicle to be planned, acquiring the stability requirement of the vehicle to be planned and the contour parameters of the vehicle to be planned, and determining the feasible parameter range of roads and bridges of feasible transportation routes; acquiring parameters of roads and bridges of each transportation road section in a transportation network topology, comparing the parameters of the roads and bridges of each transportation road section with a feasible parameter range, and selecting the transportation road section with the parameters of the roads and bridges in the feasible parameter range as a feasible transportation road section;
wherein, the step S1 specifically includes the following contents:
the method comprises the steps of firstly accessing a road information management system established by each province in a road planning subsystem, thereby meeting the requirements of acquiring road network data information, inquiring maps and the like, and selecting high-grade roads such as expressways, national roads and provincial roads according to the starting and ending points of the roads.
And then analyzing the feasibility of vehicle passing from two aspects of vehicle space physical dimension information and vehicle running stability. Firstly, acquiring parameters of roads and bridges of each road section through an information management database, such as clear width requirements, clearance requirements of tunnels, road signboard heights and the like of the roads and the bridges, and preliminarily analyzing whether the road section meets vehicle passing requirements or not by combining information such as vehicle heights, vehicle widths, vehicle lengths and the like provided by large transport enterprises; and secondly, analyzing the driving stability of the vehicle, acquiring the minimum flat curve radius, the maximum longitudinal slope and the maximum transverse slope of the route through a road and bridge information management system, and analyzing whether the route can meet the safe passing requirement of the vehicle and the like by combining the traction force of the vehicle and the binding stability of goods.
All possible transportation routes are preliminarily obtained by judging the physical size of the space and the driving stability of the vehicle through the system.
S2, acquiring load data of a vehicle to be planned by adopting a machine vision technology, comparing the vehicle standard load of a feasible transportation road section with the load data of the vehicle to be planned, and selecting the feasible transportation road section with the vehicle standard load larger than the load data of the vehicle to be planned as all safe transportation road sections of the vehicle to be planned;
in step S2, vehicle parameter information may be obtained by the machine vision system to build a vehicle load model. In the reporting process, a transportation enterprise can establish a vehicle load model through the system, so that a basis is provided for bridge safety assessment. Before actual transportation begins, a road management department can utilize the system to check actual loads and release vehicles with the difference value between the declared loads and the actual loads within a preset range; and (4) advising the vehicles with the difference value between the declared load and the actual load larger than the preset value. The method specifically comprises the following steps:
s21, identifying vehicle and tire parameters based on a machine vision technology, establishing an axle load prediction model based on an artificial neural network, inputting the identified vehicle tire parameters into the axle load prediction model, realizing axle load data of the vehicle, and calculating load data of the vehicle according to the axle load data:
as shown in fig. 2, the step S21 specifically includes the following steps:
s211, identifying the tire deformation value, the tire type and the tire pressure: tire images are acquired through high frame rate cameras erected on two sides of a road, and a Mask R-CNN convolutional neural network is trained on a tire data set in advance. In the identification process, each frame of the vehicle driving video is used as input, the Mask R-CNN automatically identifies the tire and the rim, and a segmentation Mask is generated. On the basis, the images generated by Mask R-CNN are processed by an edge detection method, so that the extraction of the tire contour and the rim contour is realized. And calculating the circle center of the rim based on the identified rim profile, and on the premise of assuming that the rim and the upper half part of the tire are not deformed, taking a horizontal straight line passing through the circle center of the rim as a symmetry axis, downwardly symmetrical the upper half part of the tire to obtain an ideal undeformed tire profile, and extracting a pixel point difference value between the ideal undeformed profile and the deformed bottom of the tire profile. The tire model, the rim diameter and the tire pressure information are identified from the tire image based on a character identification technology, and a scale factor, namely the real physical length corresponding to one pixel point in the image is calculated according to the identified rim diameter identification and the number of rim diameter pixel points in the image. And finally, multiplying the difference value of the pixel points at the bottom of the tire by a proportional factor to obtain the deformation value of the tire under the load action.
S212, building a tire load-deformation model: an Artificial Neural Network (ANN) is an information processing Network simulating the structure of a human brain, and is a nonlinear dynamical system, and although each neuron has a simple structure and cannot solve a complicated practical problem, the Neural Network constructed by the neurons is highly nonlinear and can process a lot of complicated problems. In addition, the artificial neural network also has the advantages of distributed storage, parallel processing, nonlinear mapping and the like.
The tyre has the characteristics of complex structure, complex mechanical property and the like, which causes the contact condition between the tyre and the ground to have high complexity in the running process of a vehicle. The tire theory simplification method and the fitting method are difficult to accurately predict the contact force between the tire and the ground, and are also difficult to adapt to the contact problem between tires of different models and the ground.
Therefore, the invention completes the modeling of the tire-ground contact model by using the artificial neural network, and processes the highly nonlinear problem by means of the neural network, thereby realizing the weight acquisition of the large transport vehicle. The specific implementation process comprises the following steps: a multilayer artificial neural network as shown in fig. 3 is established, each neuron in the layer is not connected, and the neurons in the layers are fully connected. The weight value of each neuron adopts random initialization, and a Sigmoid function is used as an activation function of each neuron. Training the established network on a measured data set of a tire mechanical property test, and obtaining parameter information (such as model 195/55/R16, 195 indicates that the width of the tire is 195mm, 55 indicates that the tire flatness ratio, namely the section height is 55 percent of the width), R16 indicates that the tire is a meridian tire, the diameter of a rim is 16 inches, and the corresponding network input is a vector [195, 55, 16]T) Tire pressure identification value (such as 350kPa, corresponding network input is 350), calculated flexibility value (such as 20mm, corresponding network input is 20) and number of single-side tires on each axle provided in declaration as network input (such as two tires on one side of a certain axle, corresponding network input is 2), integrating the parameters, keeping units of various physical quantities consistent with data units of a training set in training a model, and obtaining corresponding network input [195, 55, 16, 350, 20, 2 ]]TAnd outputting the predicted value of the axle load by the trained model.
The training of the neural network refers to firstly collecting measured data of a tire mechanical performance test and other tire mechanical performance test data disclosed by scholars, wherein the data comprises tire model information, a tire pressure identification value, the number of tires on one side of an axle, the magnitude of axle load and a measured tire deflection value, after the unit of the data is normalized (for example, the diameter of a rim of a tire is uniformly expressed by one inch, and the like), the data set is randomly divided into a training set, a testing set and a verification set according to the proportion of 60%, 20% and 20%, wherein the training set is used for training the neural network, the testing set is used for evaluating a model, and the verification set is used for primarily evaluating the capability of the model and correspondingly adjusting the hyper-parameters. And taking the training set as the input of the training network, carrying out forward propagation, comparing the network output value with an expected output value, and calculating to obtain an error value. The error is taken as the basis of learning, the gradient descent method is utilized to reversely propagate the error layer by layer, and the adjustment of the neuron parameter of each layer is realized, namely the learning process.
In order to obtain a more accurate prediction model, and also to accelerate the training process and avoid the training process from falling into a local minimum, appropriate network parameters need to be set before the training starts. Set up the input layer neuron number and be 6, the output layer number is 1, and the number of hidden layers is hidden layer neuron number and adjusts according to the training effect promptly, and the momentum factor gets 0.95, and the number of iterations is adjusted according to the training effect, and initial learning rate gets 0.5, and self-adaptation adjustment learning rate during training, whether check the weighted value promptly and reduced the error of output, if the error reduces, then can suitably increase learning rate, otherwise suitably reduce learning rate, and the rule of adjustment is:
wherein
Is the sum of the squared errors for the k-th iteration,
the learning rate for the kth iteration.
And continuously adjusting and optimizing the parameters, and when the relative errors of all predicted values and true values of the model on the test set are kept between-2% and 2%, considering that the obtained prediction model is relatively ideal in performance, and terminating the optimization process so as to obtain the axle weight prediction model.
S213, vehicle axle load identification: after the tire model, the tire pressure identification value and the actual deflection value are obtained based on machine vision, the information is input into a trained vehicle load recognition artificial neural network model by combining the number of tires on one side of each axle provided in declaration, and the output value of the network is used as the vehicle axle weight.
And S214, summing all the axle weights of the vehicle to obtain the total weight of the vehicle.
S22, comparing the vehicle standard load of the feasible transportation road section with the load data of the vehicle to be planned, and selecting the feasible transportation road section with the vehicle standard load larger than the load data of the vehicle to be planned as all safe transportation road sections of the vehicle to be planned:
for any feasible transportation section, the following steps are carried out:
s221, firstly, comparing load data obtained by calculation of the constructed vehicle load model with standard loads in the design specifications of the highway bridge of the feasible transportation road section:
if the vehicle load model is smaller than the standard load defined by the design specification of the highway bridge and the bridge response under the short-term effect combination of the load is smaller than a specified value, judging that the vehicle to be planned can safely pass on the feasible transportation road section, and taking the feasible transportation road section as a safe transportation road section;
if the vehicle load model is larger than the standard load defined by the design specification of the highway bridge, a pre-established limited metadata database of the bridge structure is called, the load model established based on machine vision is input into a finite element calculation model established based on ANSYS, and loading is carried out according to the actual vehicle running position, so that static and dynamic responses of the structure under the vehicle load are obtained. In order to consider material degradation, section reduction and the like of the structure caused by environmental factors, a structural resistance reduction coefficient is calculated according to the specification of 'road and bridge bearing capacity detection and assessment regulation' (JTG/T J21-2011). Therefore, the resistance of the bridge structure is reduced. After the bridge response is obtained through calculation, comparing the bridge response value with the current bridge design specification:
if the response value is not greater than the standard specified value, judging that the vehicle to be planned can safely pass through the bridge, and regarding the feasible transportation road section corresponding to the bridge as a safe transportation road section;
if the response value is larger than the specified value and not larger than 1.05 specified value, the vehicle can pass through the bridge after the bridge is reinforced, but the condition of the bridge needs to be monitored in real time during passing; the feasible transportation road section after the bridge is reinforced is a safe transportation road section;
if the response value is greater than the 1.05 specification, then no traffic is allowed.
Because the vehicle keeps in the middle, at a constant speed, at a slow speed when driving, the impact effect of the vehicle can not be considered when loading.
S3, constructing a safe transportation network topology according to all safe transportation road sections of the vehicle to be planned, and screening out an optimal transportation line from the safe transportation network topology:
and screening the routes obtained by the preliminary analysis of the road planning subsystem based on the analysis result of S2, wherein the screening condition is whether the bridge bearing capacity meets the requirement, and the screened results are all possible passing routes, but the results are possibly not optimal. Therefore, the safety factor of the bridge on the route and the length of the passing route are used as optimization indexes, and the genetic algorithm is used for completing optimization of the objective function, namely completing intelligent path planning of large piece transportation.
The method comprises the following steps of optimizing a path through a genetic algorithm under the condition of determining the starting point and the ending point of a transportation route, wherein the specific implementation process comprises the following steps:
s31, firstly, initializing parameters, wherein the initialization parameters comprise population number P, chromosome base factor N, iteration number, cross probability and mutation probability. Nodes in the road network after the checking calculation and screening of the bridge bearing capacity are then encoded by using a binary encoding scheme, for example, the node of the road network No. 5 can be encoded into 00000101. Then, an initial population is randomly generated, namely P individuals are generated, each individual is represented by a chromosome, the chromosome is composed of N gene loci, each gene locus represents a value obtained by binary coding of a node number of the route scheme, and the decoded chromosome represents a route scheme from a starting point to an end point. The numbered road networks are shown in fig. 4.
And S32, selecting a fitness function. Because the route length of each road section is a known value, a load checking calculation model of the bridge on the route is established through finite elements in the early stage, and the structural resistance and the response of the structure under the action of vehicle load are known, the total distance of a certain randomly generated road section combination can be combined
Is divided by the bridge safety reserve coefficient
As a fitness function (the principle is that the longer the route length is, the lower the bridge safety reserve is, and the lower the individual fitness is), namely the fitness function
. Wherein the safety reserve coefficient of the bridge
The maximum value of the ratio of the theoretical response value of all bridges under the vehicle load to the reduced theoretical resistance value of the bridges on the route is measured, and the value is used for measuring the safety reserve of the bridges when the vehicles pass through the bridges, namely the safety of the vehicles passing through the bridges. The larger the ratio, the lower the bridge safety reserve. Based on the selected fitness function, a fitness value is calculated for each individual in the population.
And S33, performing selection operation to determine which individuals can enter the next generation, wherein the principle is that the fitness calculated by the individuals with short route length and high bridge safety reserve is high, and the number inherited to the next generation is more. The system adopts a roulette mode, the roulette mode is a classical geometric probability model, and the selected probability is proportional to the fitness. The step of calculating the sum of individual fitness
And calculating for each individual
(ii) a Generating random number r of [0,1), calculating
To find out
If the k value is the minimum value, the kth individual in the population is selected; repeating the above process M times to obtain M next generation individuals.
S34, performing crossover operation on each individual with a certain probability, namely randomly pairing the populations and then randomly exchanging each other at a certain position of the chromosome. This operation is represented in the present system as randomly pairing the generated routes, then randomly determining the switching nodes, and interchanging the segments between the switching nodes of the two paired routes. The obtained offspring individuals after the cross operation are possible to have higher fitness.
And S35, performing mutation operation. For each individual, each genetic locus is mutated with a certain probability. After mutation operation is performed, a part of gene loci of the original individual chromosome are changed. For the binary coding scheme, mutation operation is the change of a gene on a chromosome from 1 to 0 or from 0 to 1. After the chromosome is decoded, the mutation operation is represented by changing one or more nodes in the route from the starting point to the end point into other nodes, namely, one or more road sections in the route are changed. When performing mutation operations, an appropriate mutation probability should be selected.
And S36, generating new filial generation individuals after the three operations of selection, crossing and mutation, and repeating the calculation of the fitness value and the three genetic operations until the termination condition is met. The termination condition is that the fitness of the optimal individual reaches a threshold value, or the fitness of the optimal individual and the population fitness do not rise or rise obviously, or the iteration number reaches a preset value.
Through the genetic algorithm, the path is further optimized, so that the vehicle driving distance is shortest on the basis of meeting the bridge safety, the bridge safety reserve on the selected path is higher, and the probability of damage to the bridge caused by the vehicle is reduced.
After the optimal path planning is determined, a path evaluation report can be generated through a post-processing module, and the contents of the path evaluation report comprise vehicle trafficability inspection, bridge structure bearing capacity accounting and the like, so that transportation enterprises are assisted to select optimal paths of transportation theories, and management departments are assisted to complete approval of large transportation projects.
S4, monitoring the running state of the vehicle in the optimal transportation path:
the large transport vehicle is an ultra-long, ultra-wide and special overload vehicle, and generally runs at a constant speed and slowly and centrally when running on a bridge. Since the driving state of a large transportation vehicle affects the driving safety to some extent, it is necessary to monitor the driving state of the vehicle during driving.
And realizing target identification and tracking based on the established large-scale traffic camera and a YOLOv5 algorithm and a Deepsort algorithm which are pre-trained on a tire image data set.
Because the transport vehicle is long and the field of view of the camera is limited, the vehicle is directly identified through the YOLOv5, phenomena such as disappearance of an identification frame and drastic change of the size of the identification frame in the identification process may occur, and the actual vehicle position acquisition fails or has a large error. Therefore, the system completes the task of monitoring the running state of the vehicle by identifying the tires of the vehicle.
The specific process comprises the steps of firstly establishing a tire image data set, training a YOLOv5 algorithm and a Deepsort algorithm based on the tire image data set, then acquiring camera internal parameters only related to a camera and camera external parameters related to the relative position of a camera bridge deck, and finally acquiring vehicle driving videos through traffic cameras arranged on two sides of a road in the transportation process. Because the algorithm is poor in performance when a far object is detected, a virtual detection area is set at a position 1-2 times the bridge width away from a camera, after a tire enters the detection area, the YOLOv5 algorithm automatically detects the existence of the tire, a prediction frame and the recognition confidence coefficient are given, the Deepsort realizes the tire target tracking and counting functions, namely, tires moving between frames of a video are determined to be the same tire, and IDs are distributed in sequence.
And (3) taking the middle point of the lower rectangular part of the tire identification frame as a contact point of the tire and the bridge floor, and projecting the point onto the bridge floor through the internal and external parameters of the camera after obtaining the pixel point position of the contact point, thereby obtaining the actual position of a certain tire or certain tires.
The time stamp of each tire leaving the virtual detection area is recorded during the tracking process, and the average speed of the vehicle for a certain period of time is calculated by using the wheel base between the two tires and the time difference of the tires leaving the detection area.
The scheme for acquiring the wheelbase is as follows:
vehicle axle number identification: vehicle driving videos are obtained through cameras on two sides of a road, and a Yolov5 target recognition algorithm and a Deepsort target tracking algorithm are trained on a tire image data set in advance. Taking the vehicle driving video as an algorithm input, the YOLOv5 algorithm can identify the target in each frame of the video, finally give a target prediction box, classify the target into a corresponding classification, and output the confidence of identification. The Deepsort algorithm is based on the recognition result of YOLOv5, performs feature extraction (such as apparent features and motion features) on the recognition target, calculates the matching degree of the target between the front frame and the rear frame, and finally performs data association and assigns an ID value to the target, thus realizing automatic counting of the axles.
And (3) identifying the wheel base: the method comprises the steps of obtaining a rim diameter identification based on a character recognition technology, obtaining a rim image based on Mask R-CNN segmentation, calculating the number of pixel points corresponding to the rim diameter, and calculating the relation between the lengths of a real world coordinate system and a pixel coordinate system, namely the proportional factor of 2.1.1. And marking the circle center position of the rim in real time after the segmented rim image is obtained, calculating the position difference of pixel points between different shafts based on the position difference, and multiplying the pixel point difference by a proportional factor to obtain the actual vehicle wheel base.
In addition, the invention also discloses a working platform for large piece transportation unification, and a road management department, a detection mechanism, a large piece transportation enterprise, a monitoring unit and the like can finish corresponding large piece transportation processes on the platform. The system comprises the road planning subsystem, the machine vision subsystem and the finite element calculation subsystem, and is linked to a road information management database, a bridge finite element model database and a bridge health monitoring and evaluating database. The platform realizes 'declaration, preliminary examination, safety evaluation, intelligent route optimization, transportation state monitoring' and the like by means of modernization, and greatly improves declaration, examination and approval efficiencies.
In conclusion, the vehicle load measurement and transportation route planning method and system based on machine vision can obtain more accurate vehicle parameters and load models through the camera, can monitor the vehicle running state in real time, realize automation and improve efficiency; in addition, the theoretically optimal transportation path is obtained through the analysis of a path planning system, so that the transportation cost is reduced;
in addition, an integrated large transport management platform is also established in the invention, and the highway management department, the detection mechanism, the large transport enterprise and the monitoring unit can respectively complete links such as declaration, accounting, approval, monitoring and the like on the same platform, so that the declaration, accounting, approval and monitoring efficiency can be improved.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A vehicle load measuring method based on machine vision is characterized by comprising the following steps:
collecting tire images of the left side and the right side of a vehicle to be tested, and acquiring the number of tires on each axle of the vehicle to be tested;
extracting the actual deformation, the tire model and the tire pressure information of the tire corresponding to each axle from the tire image;
inputting the actual deformation, the tire model, the tire pressure information and the number of tires on the axle of the vehicle to be tested into a trained vehicle load identification model to obtain the axle weight data of each axle of the vehicle to be tested, and calculating the load data of the vehicle to be tested based on the axle weight data of each axle; the vehicle load identification model is an artificial neural network model, the artificial neural network model comprises a plurality of layers of artificial neural networks, all neurons in each layer are not connected, all neurons in each layer are connected, the weighted value of each neuron adopts random initialization, and a Sigmoid function is used as an activation function of each neuron.
2. The machine-vision-based vehicle load measuring method of claim 1, wherein the training process of the vehicle load model is as follows:
constructing a training set of the vehicle load model, taking the training set as input of a training network, carrying out forward propagation, comparing a network output value with an expected output value, and calculating an error value; taking the error value as a learning basis, and reversely propagating the error layer by using a gradient descent method to realize the adjustment of the neuron parameter of each layer;
the learning rate is adaptively adjusted during the training of the vehicle load model, wherein the adjustment rule is as follows:
wherein the content of the first and second substances,
is the sum of the squared errors for the k-th iteration,
the learning rate for the kth iteration.
3. The machine-vision-based vehicle load measuring method according to claim 1, wherein extracting the actual deformation amount, the tire model number and the tire pressure information of the tire corresponding to each axle from the tire image comprises the following steps:
identifying the model, the rim diameter and the tire pressure information of each tire in the tire image based on a character identification algorithm; identifying each axle, the corresponding tire of each axle, the tire contour and the rim contour of each tire from the tire image based on a Mask R-CNN identification algorithm, and extracting the tire contour and the rim contour of each tire from the tire image by adopting an edge detection measuring method;
dividing the tire contour into an upper tire contour and a lower tire contour by taking a horizontal straight line passing through the circle center of the rim contour as a dividing line; calculating the difference between the pixel points of the upper tire contour and the lower tire contour as the tire deformation pixel value; and calculating a scale factor according to the identified rim diameter and the number of the rim diameter pixel points in the image, and calculating the actual deformation of the tire according to the scale factor and the tire deformation pixel value.
4. A vehicle transportation route planning method based on machine vision is characterized by comprising the following steps:
calculating load data of a vehicle to be planned by adopting the method of any one of claims 1 to 3, and determining an origin and a target of the vehicle to be planned;
determining a transportation network topology between an origin and a target, acquiring the stability requirement of a vehicle to be planned and the contour parameters of the vehicle to be planned to determine the feasible parameter range of roads and bridges of a feasible transportation line;
acquiring parameters of roads and bridges of each transportation road section in a transportation network topology, comparing the parameters of the roads and bridges of each transportation road section with a feasible parameter range, and selecting the transportation road section with the parameters of the roads and bridges in the feasible parameter range as a feasible transportation road section;
acquiring vehicle standard load of a feasible transportation road section, comparing the vehicle standard load of the feasible transportation road section with the load data of the vehicle to be planned, and selecting the feasible transportation road section with the vehicle standard load larger than the load data of the vehicle to be planned as all safe transportation road sections of the vehicle to be planned;
and constructing a safe transportation network topology according to all safe transportation road sections of the vehicle to be planned, and screening out an optimal transportation line from the safe transportation network topology.
5. The machine vision-based vehicle transportation route planning method according to claim 4, wherein the parameters of the roads, bridges comprise: the road sign board height, the minimum flat curve radius of the route, the maximum longitudinal slope and the maximum transverse slope are combined; further comprising the steps of:
identifying the wheelbase of the vehicle to be planned by adopting a machine vision technology, and constructing a load model of the vehicle to be planned by combining the axle weight of each axle of the vehicle to be planned;
when the standard load of the vehicle is smaller than the load data of the vehicle to be planned, calling pre-constructed structural data of the bridge from a finite element database, constructing a finite element calculation model of the bridge according to the structural data, and loading the load model of the vehicle to be planned on the finite element calculation model of the bridge according to the actual running position of the vehicle to be planned on the bridge, so as to obtain static and dynamic response values of the bridge under the load action of the vehicle to be planned;
comparing the static and dynamic response values with the design specification response of the bridge, if the static and dynamic response values are smaller than the design specification response values, judging that the feasible transportation road section corresponding to the bridge is a safe transportation road section, and if the static and dynamic response values are larger than the design specification response values and not larger than 1.05 times of the design specification response values, judging that the feasible transportation road section after the bridge is reinforced is the safe transportation road section; and if the static and dynamic response values are larger than the 1.05 design specification response value, judging that the feasible transportation road section corresponding to the bridge is not a safe transportation road section.
6. The machine vision based vehicle transportation route planning method according to claim 5, wherein identifying the wheelbase of the vehicle to be planned using machine vision techniques comprises the steps of:
acquiring a vehicle image simultaneously containing two axle wheels, identifying two adjacent wheels from the vehicle image based on a Mask R-CNN identification algorithm, and detecting the rim profiles of the two adjacent wheels from the vehicle image by adopting an edge detection method; marking the circle center positions of the rims of the two adjacent wheels, and calculating a pixel point value between the circle center positions of the rims of the two adjacent wheels;
for any one wheel A of two adjacent wheels, identifying the rim diameter of the wheel A from a vehicle image based on a character identification technology, and calculating a rim diameter pixel point of the wheel A in the vehicle image; calculating a scale factor according to the identified rim diameter and the number of rim diameter pixel points in the image; and multiplying the pixel point value between the circle center positions of the wheel rims of the two adjacent wheels by the scaling factor to obtain the wheel base of the two adjacent wheels.
7. The machine vision based vehicle transportation route planning method according to claim 5, characterized in that an optimal transportation route is screened out from the safe transportation network topology, comprising in particular the steps of:
constructing a target optimization function which takes a safe transportation line of a safe transportation network topology as an independent variable, and takes the minimum total distance of the safe transportation line and the maximum safety factor of the bridge;
encoding each transport node in the secure transport network topology; coding each safe transportation line between the origin and the target site based on the codes of each transportation node to obtain a coded safe transportation line set;
taking each encoded safe transportation line as a chromosome, taking the encoded transportation nodes as genes of the chromosome, and taking the target optimization function as a fitness function to construct a genetic algorithm to solve the optimal solution of the target optimization function; and taking the safe transportation line corresponding to the optimal solution as an optimal transportation line.
8. The machine vision based vehicle transportation route planning method of claim 7, further comprising the steps of:
constructing and training a wheel identification model taking YOLOv5 as a frame, acquiring a vehicle video of a vehicle in a transportation process through a camera, inputting the vehicle video into the wheel identification model to obtain vehicle videos of frames marked with wheel prediction frames, associating the prediction frames of the same wheel in the frames of the vehicle video by adopting a Deepsort algorithm, and allocating IDs to different wheels according to a sequence to realize the tracking of the vehicle;
extracting the position of a pixel point of the middle point of the lower rectangular part of the wheel prediction frame as the position of a contact point of the wheel and the bridge floor, and calculating the actual position of the vehicle according to the shooting angle and the focal length of a camera, the position of the camera and the position of the contact point;
recording the time stamp of each tire of the vehicle leaving the preset virtual detection area during the tracking process, and calculating the average speed of the vehicle by using the wheel base between two wheels and the time difference of the wheels leaving the detection area.
9. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the method of any one of claims 1 to 8 are performed when the computer program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210145462.7A CN114202572A (en) | 2022-02-17 | 2022-02-17 | Vehicle load measuring and transportation route planning method and system based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210145462.7A CN114202572A (en) | 2022-02-17 | 2022-02-17 | Vehicle load measuring and transportation route planning method and system based on machine vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114202572A true CN114202572A (en) | 2022-03-18 |
Family
ID=80645599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210145462.7A Pending CN114202572A (en) | 2022-02-17 | 2022-02-17 | Vehicle load measuring and transportation route planning method and system based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114202572A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114822029A (en) * | 2022-04-26 | 2022-07-29 | 广州大学 | Bridge deck traffic flow load space-time distribution reconstruction method, system and device |
| CN115345247A (en) * | 2022-08-22 | 2022-11-15 | 交通运输部规划研究院 | Model training and road axle load determining method, device, equipment and medium |
| CN115900906A (en) * | 2022-06-15 | 2023-04-04 | 东南大学 | Bridge dynamic weighing method based on mid-span boundary beam measuring point strain |
| CN116824524A (en) * | 2023-07-17 | 2023-09-29 | 黑龙江省卓美工程建设有限公司 | Big data flow supervision system and method based on machine vision |
| CN116843895A (en) * | 2023-05-30 | 2023-10-03 | 湖南大学 | Image recognition-based vehicle tire pressure and weight combined prediction method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110514215A (en) * | 2019-08-12 | 2019-11-29 | 贵州省质安交通工程监控检测中心有限责任公司 | A kind of Over-size transport paths planning method based on bridge structure force analysis |
| CN111144039A (en) * | 2019-12-04 | 2020-05-12 | 东南大学 | Train dynamic weighing system and weighing method based on deep learning |
| CN112836892A (en) * | 2021-02-26 | 2021-05-25 | 山东科技大学 | Multi-target vehicle distribution path determining method and system based on improved genetic algorithm |
-
2022
- 2022-02-17 CN CN202210145462.7A patent/CN114202572A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110514215A (en) * | 2019-08-12 | 2019-11-29 | 贵州省质安交通工程监控检测中心有限责任公司 | A kind of Over-size transport paths planning method based on bridge structure force analysis |
| CN111144039A (en) * | 2019-12-04 | 2020-05-12 | 东南大学 | Train dynamic weighing system and weighing method based on deep learning |
| CN112836892A (en) * | 2021-02-26 | 2021-05-25 | 山东科技大学 | Multi-target vehicle distribution path determining method and system based on improved genetic algorithm |
Non-Patent Citations (3)
| Title |
|---|
| 孔烜等: "基于图像识别轮胎变形的完全非接触式车辆称重方法", 《中国公路学报》 * |
| 罗亚波: "《作业系统调度优化理论与方法》", 30 November 2011 * |
| 赵宇兰: "《分布式数据库查询优化研究》", 31 July 2017 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114822029A (en) * | 2022-04-26 | 2022-07-29 | 广州大学 | Bridge deck traffic flow load space-time distribution reconstruction method, system and device |
| CN114822029B (en) * | 2022-04-26 | 2023-04-07 | 广州大学 | Bridge deck traffic flow load space-time distribution reconstruction method, system and device |
| CN115900906A (en) * | 2022-06-15 | 2023-04-04 | 东南大学 | Bridge dynamic weighing method based on mid-span boundary beam measuring point strain |
| CN115900906B (en) * | 2022-06-15 | 2023-07-25 | 东南大学 | Bridge dynamic weighing method based on strain of measuring point of midspan boundary beam |
| CN115345247A (en) * | 2022-08-22 | 2022-11-15 | 交通运输部规划研究院 | Model training and road axle load determining method, device, equipment and medium |
| CN116843895A (en) * | 2023-05-30 | 2023-10-03 | 湖南大学 | Image recognition-based vehicle tire pressure and weight combined prediction method and system |
| CN116824524A (en) * | 2023-07-17 | 2023-09-29 | 黑龙江省卓美工程建设有限公司 | Big data flow supervision system and method based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114202572A (en) | Vehicle load measuring and transportation route planning method and system based on machine vision | |
| US20220196459A1 (en) | Real-time vehicle overload detection method based on convolutional neural network | |
| Chakraborti et al. | A neural network and landscape metrics to propose a flexible urban growth boundary: A case study | |
| CN104346538B (en) | Earthquake hazard assessment method based on three kinds of the condition of a disaster factor control | |
| CN113989230A (en) | Improved YOLOv 4-based road pavement disease detection method | |
| Akgüngör et al. | An artificial intelligent approach to traffic accident estimation: Model development and application | |
| CN102564546A (en) | Method and system for detecting overload and unlawful measurement of vehicle | |
| CN113255811A (en) | Beam bridge multi-source heterogeneous data fusion decision-making system based on BIM | |
| CN114372117A (en) | Multi-scale aggregation mode analysis method for complex traffic network | |
| CN106970986A (en) | Urban waterlogging influence degree method for digging and system based on deep learning | |
| CN103150774A (en) | System and method for identifying vehicles of highway green channel | |
| CN113918538B (en) | New road maintenance data migration system based on artificial neural network | |
| Shen et al. | Identification of accident blackspots on rural roads using grid clustering and principal component clustering | |
| CN111126868A (en) | Road traffic accident occurrence risk determination method and system | |
| CN113393665A (en) | Planning method for dangerous goods transportation path under uncertain time-varying road network | |
| Astarita et al. | Developing new hybrid grey wolf optimization-based artificial neural network for predicting road crash severity | |
| Ge et al. | Intelligent simulation method of bridge traffic flow load combining machine vision and weigh-in-motion monitoring | |
| CN116448224A (en) | Bridge dynamic weighing and grading alarm method based on visual strain monitoring | |
| Lei et al. | Condition level deteriorations modeling of RC beam bridges with U-Net convolutional neural networks | |
| Habibzadeh et al. | Application of artificial neural network approaches for predicting accident severity on rural roads (case study: tehran-qom and tehran-saveh rural roads) | |
| Aloisio et al. | Damage classification after the 2009 L’Aquila earthquake using multinomial logistic regression and neural networks | |
| CN114118539A (en) | Highway accurate maintenance decision-making method based on microscopic unit performance prediction | |
| Menguc et al. | A Data Driven Approach to Forecasting Traffic Speed Classes Using Extreme Gradient Boosting Algorithm and Graph Theory | |
| CN117133099A (en) | Automatic monitoring system for disaster of expressway high slope | |
| CN102880881A (en) | Method for identifying car type on basis of binary support vector machines and genetic algorithm |
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 |
Application publication date: 20220318 |
|
| RJ01 | Rejection of invention patent application after publication |