CN102855499A - Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines - Google Patents
Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines Download PDFInfo
- Publication number
- CN102855499A CN102855499A CN2012103024691A CN201210302469A CN102855499A CN 102855499 A CN102855499 A CN 102855499A CN 2012103024691 A CN2012103024691 A CN 2012103024691A CN 201210302469 A CN201210302469 A CN 201210302469A CN 102855499 A CN102855499 A CN 102855499A
- Authority
- CN
- China
- Prior art keywords
- roofbolt
- bar
- data pipe
- information data
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention discloses a method for determining railway contact net main column identification information based on combination of multiple clue data pipelines. The method mainly comprises the following steps of: 1) building the data pipelines; 2) generating data pipeline intersection points; 3) calculating a column number and a kilometer post by combining the multiple clue data pipelines; and 4) updating the data pipelines. By the method, positions of images, such as shaded and polluted column numbers or kilometer posts, which cannot be identified visually can be precisely calculated by combining the multiple clue data pipelines by fully using a middle result of image identification of the contact net; and the method is also suitable for inspection and correction of column number and kilometer post image identification results of inspection images in an electrical railway contact net. The calculation method based on multiple clue data pipelines can be also applied to tasks similar to application of the type.
Description
Technical field
The present invention relates to data fusion analysis and Computer Recognition Technology field, specifically a kind of contact net roofbolt that merges based on the multi thread data pipe number and kilometer post projectional technique.
Background technology
Electrification railway contact net is the special circuit that sets up for the bullet train power supply.In order to guarantee the safe operation of bullet train, contact net need to often be patrolled and examined, and the present equipment that the high speed passenger dedicated railway contact net equipment of 350km/h is maked an inspection tour online is first one-step forming.Yet, the various potential safety hazards of contact net effectively being detected, inspection device need record the multitude of video data, and in the face of the video data of patrolling and examining of magnanimity, artificial interpretation workload is large, efficient is low, reliability is difficult to ensure if only depend on.How contact net is patrolled and examined image king-rod number and kilometer post carries out Intelligent Recognition, make up retrieval and the management system of each roofbolt image of electrification railway contact net, reduce the picture amount that needs interpretation, efficient is patrolled and examined in raising and managerial ability becomes exigence.On the one hand, the roofbolt identification information really surely for the contact net potential faults automatically identification the signal (such as roofbolt number and kilometer post) of physical location is provided, on the other hand, can gather the image Rapid Establishment or upgrade roofbolt information management archives for Daily Round Check.Yet, part roofbolt upper boom number or kilometer post zone exist blocks, pollutes or the severe jamming such as damaged, only utilizing visual information that these roofbolts are identified its reliability can't ensure, this is long to circuit, the identification of the bar of the electric railway wide, that environment is various that distributes number and kilometer post has determined to bring huge difficulty.
Summary of the invention
In view of the visual identity blind area that in contact net is maked an inspection tour image, has bar number and kilometer post, the object of the invention is to propose a kind of multi thread data pipe integration technology, fully utilize various supplementarys image recognition intermediate result is carried out verification and correction, the bar of vision dead zone number and kilometer post are rationally calculated.
To achieve these goals, the present invention adopts following means:
Definite method of the railway contact line roofbolt identification information that merges based on the multi thread data pipe is patrolled and examined video data to electrification railway contact net and is processed, and with the railway contact line roofbolt that obtains ad-hoc location number and kilometer label, may further comprise the steps:
(1) data pipe is set up, and sets up artificial discriminative information data pipe, circuit roofbolt order information data pipe, image bar identifying information data pipe, image kilometer post identification information data pipeline and image roofbolt location information data pipeline;
(2) the data pipe joint generates, and generates the data pipe joint that has three classes and have different priorities, comprising: the artificial pre-judgement roofbolt information pipeline that priority is the highest and the joint of circuit roofbolt order information data pipe; The joint of the image bar identifying information that priority is taken second place, kilometer post identification information data pipeline and circuit roofbolt order information data pipe; The joint of the image bar identifying information data pipe that priority is minimum and circuit roofbolt order information data pipe;
(3) the multi thread data pipe merges and to carry out bar number and kilometer post is calculated;
(4) data pipe upgrades.
Adopt the inventive method, can effectively calculate rod number or the positions of the unrecognizable image of vision in sequence such as kilometer post is blocked, pollution.Simultaneously, the inventive method also can expand in the task of similar application, namely has mode identification procedure and fixing pattern order in the task.As: the identification of highway kilometer post etc.
Description of drawings
The tour image synoptic diagram that Fig. 1 electrification railway contact net bar number is blocked
Fig. 2 data pipe is set up synoptic diagram
Fig. 3 roofbolt positional information matched curve synoptic diagram
Fig. 4 data pipe joint product process figure, wherein, Fig. 4 (a), Fig. 4 (a) are different with Fig. 4 (c)
The product process figure of data pipe joint
Fig. 5 multi thread data pipe projectional technique process flow diagram.
Embodiment
The electrification railway contact net roofbolt number and the kilometer post projectional technique that the present invention is based on the multi thread data pipe mainly are to have utilized the intermediate result information in the image recognition, suppose that at two strategy is calculated in the employing classification between the correct pipeline joint, thus the rational reckoning of output result.
For ease of better understanding detailed process of the present invention, the below is described in further detail the implementation procedure of the inventive method:
The data pipe that makes up comprises: roofbolt location information data pipeline, image bar identifying information data pipe, image kilometer post identification information data pipeline, artificial pre-judgement roofbolt information data pipeline and circuit roofbolt order information data pipe in the image, and the key step of its enforcement is:
Step 1: the foundation of data pipe.Each roofbolt is on time dimension in the image sequence, can enter from image one end all the time, the other end withdraws from, and along with its position gradually near video camera, roofbolt shared zone in image progressively becomes large.According to these characteristics, the present invention detects image sequence king-rod columnar object, and horizontal coordinate, roofbolt width and the frame number of record roofbolt in image set up roofbolt location information data pipeline; Utilize the recognition result of image sequence king-rod number, comprise frame number, bar number, bar recognition credibility (0～1), set up image bar identifying information data pipe; In like manner, utilize the recognition result to kilometer post in the image sequence, comprise frame number, kilometer label, kilometer post identification confidence level (0～1), set up image kilometer post identification information data pipeline; Before whole sequence was identified automatically, the inventive method allowed the frame of any amount in the sequence is manually adjudicated, and discriminative information comprises frame number, bar number, place kilometer post, sets up artificial pre-judgement roofbolt information data pipeline according to this information; , comprising by manually providing in advance for its fixing roofbolt order information of certain electric gasification rail track: circuit number, bar number, affiliated kilometer post, roofbolt have or not kilometer post, these information structures the roofbolt order information data pipe corresponding with this circuit.
Step 2: generated data pipeline joint.The inventive method merges a plurality of data pipe information calculates respective nodes in the front data pipe that needs matching image bar number and kilometer post identification information, artificial pre-judgement roofbolt information and circuit roofbolt order information to roofbolt number and kilometer post, these nodes have identical bar number and kilometer label, i.e. " the data pipe joint " of indication among the present invention.Image automatic identification is begun the roofbolt Information generation primary sources pipeline joint (ID of the front artificial judgement of scanning in circuit roofbolt order information data pipe
1, KM
1, Ind
1), joint information comprises bar number, kilometer label, frame number; In automatic identifying, when bar number and kilometer post is all identified successfully and confidence level respectively greater than threshold value T
ε pAnd T
ε kThe time, in circuit roofbolt order information data pipe take kilometer post and bar number as condition scanning, if having this bar number and kilometer post, then generate secondary sources pipeline joint (ID
2, KM
2, Ind
2), otherwise continue next identification; When the confidence level of only having image bar number identification greater than a threshold value T
ε, inquiry this bar number in circuit roofbolt order information data pipe correspondent section zone then is if exist this bar number then to generate the 3rd class data pipe joint (ID
3, KM
3, Ind
3).
Step 3: utilize the multi thread data pipe to merge to carry out bar number and kilometer post to calculate.Between the adjacent joint that described step 2 obtains, need to be in conjunction with each data pipe information, therebetween bar number is calculated, to obtain each bar number corresponding frame number.The inventive method adopts a kind of classification projectional technique that the bar between the joint number is calculated.The first order calculates and to search in the image bar identifying information pipeline, bar number identification is arranged but confidence level less than T
εAnd the node that bar number exists in that section circuit roofbolt order information is inserted into the kilometer label of this bar correspondence in the frame number of correspondence image in its image bar identifying information data pipe, the bar identified number and the circuit roofbolt order information data pipe between two joints.Calculate the second level is to utilize roofbolt positional information pipeline in the image, at first use first order curve match roofbolt change in location curve, then image recognition algorithm is detected that bar exists but the frame of failing to identify its bar number or kilometer post, according to its bar that should mate of the bar dead reckoning of identifying number and a kilometer label, and the frame number of correspondence, bar number and kilometer label inserted between two joints.When the bar quantity of extrapolating when the second level is not equal to actual quantity, be that roofbolt exists undetected or flase drop, then calculate correction, modification method is: at first should locate the second level and calculate result's deletion, the frame number of then dividing equally two joints is to corresponding interval bar in the circuit roofbolt order information pipeline number.
Step 4: data pipe upgrades.Reckoning process of the present invention can be carried out synchronously with image recognition processes, in step 3 is finished two joint among roofbolts number and is patrolled and examined video after the coupling of corresponding frame number, search for next joint, repeating step 2 is to step 4, until handle the pipeline data between all adjacent joints.
Fig. 1 is the roofbolt that a typical bar number is blocked, and it has been generally acknowledged that this type of roofbolt is infeasible in visual identity.Fig. 2 is the synoptic diagram that data pipe is set up among the present invention, and wherein artificial discriminative information data pipe and circuit roofbolt order information data pipe were set up before image automatic identification, and other three data pipelines all are to set up in image recognition processes.Roofbolt location information data pipeline has recorded horizontal level and the roofbolt width of roofbolt in each frame.For error and the true ruuning situation of reduction train of removing the location of roofbolt in the image recognition, the present invention utilizes the relation of each roofbolt of first-order linear system modelling horizontal level and time shaft in image, such as formula (1).
Wherein, t
iRepresent frame number, X
iBe the roofbolt horizontal level that image recognition obtains, N is that a bar is supposed the frame number that occurs in picture, can be similar to the frame number between twice saltus step in the reality.Parameter a is relevant with train speed, parameter b and this linear system initial, to finish frame number relevant.Utilize least square method to be easy to solve a and b, the curve of match as shown in Figure 3, every the oblique line section that wherein consists of whole curve has reflected that a roofbolt enters video pictures to the process of leaving video pictures, and this curve will be calculated for the second level of calculating algorithm.
(a), (b), (c) are the product process figure of various data pipe joints among Fig. 4.The joint that among the figure (a) is artificial pre-judgement roofbolt information and circuit roofbolt order information data pipe generates, and priority was the highest when this joint was used for calculating; Figure (b) carries out at image bar number and kilometer post identification result, when bar number and kilometer post are all identified successfully, and can correspondence be arranged with circuit roofbolt order data, then thinks a joint, and its priority is taken second place; Figure (c) is after the image bar number identify successfully, when the confidence level of exporting during greater than certain threshold value, if in the inferior order sequenced data of circuit roofbolt correspondence is arranged, then generates a joint, and its priority is minimum.In the real process, also do not satisfy confidence level greater than the identification bar of threshold value number when reaching certain frame number, the confidence level threshold value will reduce by half.Occur in the joint reality that image bar identifying information data pipe and circuit order information data pipe form at most, usefulness at most.
Fig. 5 is the process flow diagram that utilizes the multi thread data pipe to calculate.Suppose that having the frame number of two data pipeline joints is n1 and n2, and extrapolating by circuit roofbolt order information has N bar number in n1 to the n2 scope, first order reckoning process is to detect to be between n1 and the n2 to have identification but the not high bar of confidence level number, specific practice is: the ducted bar of image recognition information data is sought in circuit roofbolt order information data pipe number one by one corresponding, if the bar of identification number exists and this bar number do not occurring before this then preserve this bar number, frame number and kilometer post under it.Because bar is number undetected or the existence of flase drop, the bar number M that the first order is extrapolated is usually less than N, second level reckoning process is to carry out on the basis of the first order, if infer by route lever column information data pipe between two bars that the first order is calculated number to have undetected roofbolt, then carry out the second level and calculate.In the present embodiment, provide two kinds of secondarys to calculate algorithm, a kind of is equal point-score, and another kind is linear fitting.All point-score be take train at the uniform velocity for hypothesis, the first order is calculated the bar of existence between the bar number number is divided equally; Linear fitting calculates it then is the frame that carries out accurate backstay place in conjunction with the curve of this interval roofbolt location information data pipeline and its match.Such as Fig. 5, at first to scanning at this interval every frame of roofbolt location information data, when satisfy frame number and roofbolt position: (1) is in such as the negative slope section of Fig. 3 matched curve and in the centre position; (2) and upper secondary calculate that bar interframe exists saltus step interval, namely belong to different roofbolts and enter process, then preserve this bar number, frame number and kilometer post under it.After if linear fitting has been calculated, when finding that by contrast circuit roofbolt information data pipeline bar quantity is not equal to actual bar quantity, then revise: at first should locate linear fitting and calculate result's deletion, the frame number of then dividing equally two joints is to corresponding interval bar in the circuit roofbolt order information pipeline number.
After two-stage is calculated and finished roofbolt positional information pipeline, image bar identifying information pipeline, image kilometer post identification message tube road in the image will be updated to the information of only having terminal joint, continuation along with image recognition, before next joint occurring, recognition result will enter this three data pipelines, thereby constantly set up, calculate, upgrade, until end of identification.
Claims (3)
1. definite method of the railway contact line roofbolt identification information that merges based on the multi thread data pipe is patrolled and examined video data to electrification railway contact net and is processed, and with the railway contact line roofbolt that obtains ad-hoc location number and kilometer label, may further comprise the steps:
(1) data pipe is set up, and sets up artificial discriminative information data pipe, circuit roofbolt order information data pipe, image bar identifying information data pipe, image kilometer post identification information data pipeline and image roofbolt location information data pipeline;
(2) the data pipe joint generates, and generates the data pipe joint that has three classes and have different priorities, comprising: the artificial pre-judgement roofbolt information pipeline that priority is the highest and the joint of circuit roofbolt order information data pipe; The joint of the image bar identifying information that priority is taken second place, kilometer post identification information data pipeline and circuit roofbolt order information data pipe; The joint of the image bar identifying information data pipe that priority is minimum and circuit roofbolt order information data pipe;
(3) the multi thread data pipe merges and to carry out bar number and kilometer post is calculated;
(4) data pipe upgrades.
2. definite method of the described railway contact line roofbolt identification information that merges based on the multi thread data pipe according to claim 1 is characterized in that, described artificial discriminative information data pipe is comprised of frame number, bar number, place kilometer post; Described circuit roofbolt order information data pipe by circuit number, bar number, affiliated kilometer post, have or not the kilometer post sign to form; Described image bar identifying information data pipe is comprised of frame number, bar number, bar confidence level; Described image kilometer post identification information data pipeline is comprised of frame number, kilometer post, kilometer post confidence level; Described image roofbolt location information data pipeline is comprised of horizontal coordinate, roofbolt width and the frame number of roofbolt place image.
3. definite method of the described railway contact line roofbolt identification information that merges based on the multi thread data pipe according to claim 1 is characterized in that step (3) specifically finished by following steps:
(1) in the image recognition processes, whether monitoring two data pipeline joints occur;
(2) when detecting two adjacent joints, at first carrying out the first order between two joints calculates, be about to the ducted bar of image recognition information data seek in the circuit roofbolt order information data pipe number one by one corresponding, if having this bar number and this bar number do not occurring before this then preserve this bar number, frame number and kilometer post under it;
(3) after the first order has been calculated, check the first order calculates between the bar number whether also have detected bar;
When (4) having bar between the first order is calculated bar number, be at the uniform velocity if manually set train, then the second level is calculated directly by equal point-score bar number is divided equally between two first order are calculated bars number; If inartificial setting at the uniform velocity, then the second level calculates it is the frame that carries out accurate backstay place in conjunction with the curve of this interval roofbolt location information data pipeline and its match;
(5) calculate it is not when calculating by equal point-score when the second level, and the second level calculates that bar quantity is less than the bar quantity between two nodes calculating in one-level, then delete herein secondary and calculate the result, again press and divide equally rule and calculate that one-level calculates the bar number between the node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103024691A CN102855499A (en) | 2012-08-23 | 2012-08-23 | Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103024691A CN102855499A (en) | 2012-08-23 | 2012-08-23 | Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102855499A true CN102855499A (en) | 2013-01-02 |
Family
ID=47402075
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012103024691A Pending CN102855499A (en) | 2012-08-23 | 2012-08-23 | Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102855499A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104298993A (en) * | 2014-01-09 | 2015-01-21 | 郑州金惠计算机***工程有限公司 | Pole number positioning and identification method suitable for railway line in complex scene |
| CN104615974A (en) * | 2015-01-15 | 2015-05-13 | 成都交大光芒科技股份有限公司 | Continuous supporting pole number plate image recognizing method based on tracking algorithm |
| CN110378251A (en) * | 2019-06-28 | 2019-10-25 | 湖南华菱涟源钢铁有限公司 | Control method, device and the readable storage medium storing program for executing of train weighing apparatus department scale metering system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202098297U (en) * | 2011-06-24 | 2012-01-04 | 成都唐源电气有限责任公司 | Automatic inspecting device of key overhead contact network components |
| CN102538762A (en) * | 2012-01-10 | 2012-07-04 | 广州科易光电技术有限公司 | Online inspection device of high-speed railway contact network and inspection method of online inspection device as well as high-speed rail contact network detection system formed by online inspection device |
-
2012
- 2012-08-23 CN CN2012103024691A patent/CN102855499A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202098297U (en) * | 2011-06-24 | 2012-01-04 | 成都唐源电气有限责任公司 | Automatic inspecting device of key overhead contact network components |
| CN102538762A (en) * | 2012-01-10 | 2012-07-04 | 广州科易光电技术有限公司 | Online inspection device of high-speed railway contact network and inspection method of online inspection device as well as high-speed rail contact network detection system formed by online inspection device |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104298993A (en) * | 2014-01-09 | 2015-01-21 | 郑州金惠计算机***工程有限公司 | Pole number positioning and identification method suitable for railway line in complex scene |
| CN104298993B (en) * | 2014-01-09 | 2018-11-09 | 郑州金惠计算机***工程有限公司 | A kind of bar number positioning and recognition methods suitable under complex scene along track |
| CN104615974A (en) * | 2015-01-15 | 2015-05-13 | 成都交大光芒科技股份有限公司 | Continuous supporting pole number plate image recognizing method based on tracking algorithm |
| CN104615974B (en) * | 2015-01-15 | 2017-10-13 | 成都交大光芒科技股份有限公司 | Continuous pillar number plate image-recognizing method based on track algorithm |
| CN110378251A (en) * | 2019-06-28 | 2019-10-25 | 湖南华菱涟源钢铁有限公司 | Control method, device and the readable storage medium storing program for executing of train weighing apparatus department scale metering system |
| CN110378251B (en) * | 2019-06-28 | 2021-04-27 | 湖南华菱涟源钢铁有限公司 | Control method and device for weighing system of train weighing department and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110364008B (en) | Road condition determining method and device, computer equipment and storage medium | |
| KR101926561B1 (en) | Road crack detection apparatus of patch unit and method thereof, and computer program for executing the same | |
| CN103675868B (en) | For using the method and system of vision data determination object's position | |
| CN102116639B (en) | Method and device for automatically checking traffic limit information of electronic map | |
| CN102622886B (en) | Video-based method for detecting violation lane-changing incident of vehicle | |
| CN105608431A (en) | Vehicle number and traffic flow speed based highway congestion detection method | |
| Al-Absi et al. | Vision-based automated parking system | |
| CN110688922A (en) | Deep learning-based traffic jam detection system and detection method | |
| CN105513349A (en) | Double-perspective learning-based mountainous area highway vehicle event detection method | |
| CN102538762A (en) | Online inspection device of high-speed railway contact network and inspection method of online inspection device as well as high-speed rail contact network detection system formed by online inspection device | |
| CN108154146A (en) | A kind of car tracing method based on image identification | |
| CN111339905A (en) | CIM well lid state visual detection system based on deep learning and multi-view angle | |
| CN112883948B (en) | Semantic segmentation and edge detection model building and guardrail abnormity monitoring method | |
| Wang et al. | Realtime wide-area vehicle trajectory tracking using millimeter-wave radar sensors and the open TJRD TS dataset | |
| CN110781266A (en) | Urban perception data processing method based on time-space causal relationship | |
| CN109029372A (en) | A kind of detection vehicle localization method and device | |
| Huang et al. | Rapid surface damage detection equipment for subway tunnels based on machine vision system | |
| CN106128126A (en) | Plane cognition technology is utilized to reduce the method and system meeting red light number of times rush hour | |
| CN109029373A (en) | The synthesized positioning method and device of vehicle are detected in a kind of tunnel | |
| CN102855499A (en) | Method for determining railway contact net main column identification information based on combination of multiple clue data pipelines | |
| CN114248819A (en) | Railway intrusion foreign matter unmanned aerial vehicle detection method, device and system based on deep learning | |
| CN109029374A (en) | It is a kind of to analyze the method and device positioned to measuring car using visual pattern | |
| CN116128360A (en) | Road traffic congestion level evaluation method and device, electronic equipment and storage medium | |
| Bai et al. | Deep cascaded neural networks for automatic detection of structural damage and cracks from images | |
| CN114633774A (en) | Rail transit fault detection system based on artificial intelligence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130102 |