CN109740410B - Train set fault identification method and device without preset template - Google Patents
Train set fault identification method and device without preset template Download PDFInfo
- Publication number
- CN109740410B CN109740410B CN201811320495.0A CN201811320495A CN109740410B CN 109740410 B CN109740410 B CN 109740410B CN 201811320495 A CN201811320495 A CN 201811320495A CN 109740410 B CN109740410 B CN 109740410B
- Authority
- CN
- China
- Prior art keywords
- image
- train
- train set
- images
- target
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a train set fault identification method and a train set fault identification device without a preset template, wherein the method comprises the following steps: identifying the head and the tail of the train set according to the acquired images of the train set, and removing images in front of the head and images behind the tail in the acquired images of the train set to obtain images of the complete train set; identifying a carriage junction of the train set according to the image of the complete train set; splicing the images of the complete train set according to the connection position of the carriages to obtain a complete image of each carriage of the train set; and performing characteristic matching on the target compartment image and the reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set. According to the train set fault identification method and device without the preset template, provided by the embodiment of the invention, the real-time analysis and identification of the fault are completed on the premise of not needing a historical vehicle template library by utilizing the currently acquired train set image.
Description
Technical Field
The invention relates to the field of fault identification, in particular to a train set fault identification method and device without a preset template.
Background
In order to meet the increasing transportation requirements of people, the running speed of the railway train is gradually increased, the density is also continuously increased, and the reliability requirement of railway transportation production on the train running is higher and higher. The motor train unit train is the most important mobile equipment for completing the railway high-speed transportation task, the motor train unit has high running density, long operating mileage, complex application environment, high running speed and dense arrangement of running charts, and any tiny and fine fault can cause major accidents in a high-speed running state, thereby influencing the running safety of the high-speed railway, causing transportation interruption and line blockage, causing great loss to national economy and ensuring the important fault identification accuracy of the motor train unit.
However, most of the existing automatic motor train unit fault identification methods are based on a large number of preset template libraries, corresponding historical templates of the existing automatic motor train unit fault identification methods are retrieved according to train carriage numbers, and feature matching is carried out, so that whether potential faults exist or not is judged.
The existing motor train unit fault identification method has the following problems:
firstly, due to the fact that the preset template library has external interference factors such as illumination, driving speed and the like due to different shooting time, although the preset template library is the same train, the acquired images are greatly different, and therefore the high false alarm rate is caused.
Secondly, the position of some parts can be changed when the train is repaired, so that a template library needs to be added to a great extent, otherwise, a great amount of false alarms occur.
Disclosure of Invention
In order to overcome the technical defects, the embodiment of the invention provides a train set fault identification method and device without a preset template.
In a first aspect, an embodiment of the present invention provides a train set fault identification method without a preset template, including:
identifying the head and the tail of the train set according to the acquired images of the train set, and removing images in front of the head and images behind the tail in the acquired images of the train set to obtain images of the complete train set;
identifying a carriage junction of the train group according to the complete train group image;
splicing the images of the complete train set according to the carriage connection part to obtain a complete image of each carriage of the train set;
and performing characteristic matching on the target compartment image and a reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
In a second aspect, an embodiment of the present invention provides a train set fault identification apparatus without a preset template, including:
the first processing module is used for identifying the head and the tail of the train set according to the collected images of the train set, and removing images in front of the head and images behind the tail in the collected images of the train set to obtain images of the whole train set;
the second processing module is used for identifying the carriage connection part of the train set according to the complete train set image;
the splicing module is used for splicing the images of the complete train set according to the carriage joints to obtain a complete image of each carriage of the train set;
and the identification module is used for carrying out feature matching on the target compartment image and the reference compartment image and identifying the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
In a third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, where the processor and the memory complete communication with each other through a bus; the memory stores program instructions executable by the processor, the processor being capable of performing the method of the first aspect when invoked by the processor.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the train set fault identification method according to the first aspect without requiring a preset template.
According to the train set fault identification method and device without the preset template, provided by the embodiment of the invention, a train set head and tail template library is not required to be preset, and a train type is not required to be identified, but the currently acquired train image is utilized, the train set head and tail are firstly identified, then the carriage connection part of the train set is identified according to the correlation between the images and the characteristics of the image, and template matching is carried out with other carriage images of the train set to be detected, fault identification is realized on the premise of not searching a historical train template library according to the train type, and real-time analysis and automatic early warning of the visible structure abnormal condition of the running motor train unit are completed. The method has the advantages that the current collected image of the same train is utilized to carry out fault recognition, so that the influence of external interference factors such as illumination, driving speed and the like is avoided to a great extent, the influence of manual repair modes such as position adjustment of parts and the like is avoided, the problems of high false alarm rate, high missing report rate and the like caused by external factors such as illumination, train running speed and the like in the current fault recognition are effectively solved, the efficiency of fault recognition and maintenance operation of the motor train unit is improved, and the discovery and early warning capability of hidden faults of the motor train unit in the running process is improved.
Drawings
Fig. 1 is a schematic flowchart of a train set fault identification method without a preset template according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a vehicle head and tail identification algorithm of a gaussian mixture model according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of a car junction identification algorithm for detecting image symmetry according to an embodiment of the present invention;
fig. 4 is a schematic flowchart of a train set fault identification method without a preset template according to another embodiment of the present invention;
fig. 5 is a schematic structural diagram of a train set fault identification device without a preset template according to an embodiment of the present invention;
fig. 6 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Most of existing train set fault identification methods are based on a large number of preset template libraries, corresponding historical templates of train set fault identification methods are searched according to train carriage numbers, and feature matching is conducted, so that whether potential faults exist or not is judged. Due to the fact that the preset template library has different shooting time, the collected images are greatly distinguished due to external interference factors such as illumination and driving speed, the false alarm rate is high, and meanwhile, the template library needs to be newly added due to the change of positions of train parts. In order to solve the above problems, embodiments of the present invention provide a train set fault identification method, which directly collects a train set image to be detected on site without presetting a template library, and performs fault identification according to the train set image collected on site, thereby avoiding the influence of external factors such as illumination, traveling, position change of parts, and the like.
Fig. 1 is a schematic flow chart of a train set fault identification method without a preset template according to an embodiment of the present invention, as shown in fig. 1, including:
and 14, performing characteristic matching on the target compartment image and a reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
The train set fault identification method without the preset template provided by the embodiment of the invention is different from the existing train set fault identification method in that the existing train set fault identification method is based on a large number of head and tail preset template libraries, before fault identification, images of various different types of train sets need to be acquired, the train heads and the train tails of the train sets to be detected are detected according to the models of the train sets to be detected, corresponding historical templates are searched according to the carriage numbers of the train sets, therefore, the fault position is further determined, the train set fault identification method without the preset template provided by the embodiment of the invention does not need any historical template image, only needs to automatically identify the positions of the head and the tail of the train set according to the currently acquired train set image, removing images before the appearance of the train head and images after the appearance of the train tail, which do not belong to train images, in the acquired images; and identifying the carriage joints of the train group according to the modified complete train group image, and splicing the collected images according to the positions of the carriage joints, so that the collected images are spliced and divided according to the carriage serial numbers to form a complete image of each carriage.
The specific operation is that for a group of train groups to be detected, when the train group is detected to appear, the current images of the train groups are collected, the collected images are used as each frame of a video image, the first image and the last image are respectively processed, the images where the train head appears and the images where the train tail appears are found, then the specific position of the train head in the images where the train head appears and the specific position of the train tail in the images where the train tail appears are judged, and the specific areas of the train head and the train tail are extracted. After the position of the head of the train and the position of the tail of the train are determined, images in front of the head of the train and images behind the tail of the train in the images of the train group are removed, and the purpose of the step is to remove the images which are irrelevant to the train in the images of the train group to obtain the images of the whole train group.
After the train head and the train tail of the train set are identified, the carriage connection part of the train set needs to be further identified. The train set has certain characteristics, and the identification of the carriage connection part can be realized according to the characteristics of the train set. Taking a motor train unit as an example, the motor train unit is composed of at least two carriages with driving force and a plurality of carriages without traction force, wherein the two carriages with the driving force are two power carriages which are reversely symmetrical, namely a head and a tail of the motor train unit, the plurality of carriages without the traction force are carriages in the middle of the head and the tail, and the carriages in the middle are symmetrical to each other. The middle of any two carriages is the carriage joint. In the previous step, the image part irrelevant to the train in the train group image is removed to obtain the complete train group image. The color images of the carriages, the carriage joints and the background of the motor train unit are different, the carriages of the motor train unit can be firstly identified on the image of the complete train set, for other areas, whether the images on two sides of the area are symmetrical or not can be seen in order to distinguish the carriage joints from the background image, if the images are symmetrical, the area is the carriage joints, and if not, the area is the background image. And after the carriage joint of each image is identified, splicing the images. And if the images do not have the carriage joints, directly splicing the images, if the images have the carriage joints, splicing the left half section of the images as the carriage of the section, and splicing the right half section of the images as the starting point of the next carriage, so that the splicing of each carriage of the motor train unit is completed.
After the train set head and tail and the carriage junction of the train set are identified, fault identification is needed, and the fault identification in the embodiment of the invention still adopts the train set image acquired on site instead of a preset template. Specifically, for a certain compartment with a possible fault, the corresponding spliced image is taken out and used as a target compartment image, images corresponding to other compartments of the train set are used as reference compartment images, feature matching is carried out on the target compartment image and the reference compartment images to obtain a matching value, and the matching value is compared with a preset threshold value. If the matching value is higher than the preset threshold value, the target compartment is indicated to have no fault; otherwise, the target compartment is indicated to be in fault, and the fault position is output according to the matching result.
According to the train set fault identification method without the preset template, provided by the embodiment of the invention, a train head and tail template library is not required to be preset, and the train type is not required to be identified, but the currently acquired image of the train is utilized, the train head and tail of the train set are firstly identified, then the carriage connection part of the train set is identified according to the correlation between the images and the characteristics of the image, and then the carriage connection part is subjected to template matching with other carriage images of the train set to be detected, the fault identification is realized on the premise of not searching a historical vehicle template library according to the train type, and the real-time analysis and the automatic early warning of the visual structure abnormal condition of the running motor train unit are completed. The method has the advantages that the current collected image of the same train is utilized to carry out fault recognition, so that the influence of external interference factors such as illumination, driving speed and the like is avoided to a great extent, the influence of manual repair modes such as position adjustment of parts and the like is avoided, the problems of high false alarm rate, high missing report rate and the like caused by external factors such as illumination, train running speed and the like in the current fault recognition are effectively solved, the efficiency of fault recognition and maintenance operation of the motor train unit is improved, and the discovery and early warning capability of hidden faults of the motor train unit in the running process is improved.
On the basis of the embodiment, the train head and the train tail of the train set are obtained by identifying according to a train head and train tail identification algorithm of a Gaussian mixture model.
The vehicle head and tail identification algorithm of the Gaussian mixture model specifically comprises the following steps:
based on a Gaussian mixture model, learning and modeling are carried out on image data without train information to obtain a background image;
performing background difference on the train set image by combining the background image to obtain an image of the train head and an image of the train tail;
and determining the specific position of the vehicle head in the image where the vehicle head appears, extracting a vehicle head image area, determining the specific position of the vehicle tail in the image where the vehicle tail appears, and extracting a vehicle tail image area.
The gaussian model is a model formed based on a gaussian probability density function, which accurately quantizes an object using the gaussian probability density function (normal distribution curve) and decomposes the object into a plurality of objects. In the detection and extraction of the moving target, a background target is important for the identification and tracking of the target. Modeling is an important link of background target extraction.
First, the concept of background and foreground is to be mentioned, where foreground means that any meaningful moving object is the foreground under the condition that the background is assumed to be static. The basic idea of modeling is to extract the foreground from the current frame with the aim of bringing the background closer to the background of the current video frame. The background is updated by performing weighted average on the current frame and the current background frame in the video sequence, but due to the sudden change of illumination and the influence of other external environments, the background after general modeling is not very clean and clear, and a Gaussian Mixture Model (hereinafter referred to as GMM) is one of the most successful methods for modeling, and meanwhile, the GMM can be used for monitoring video indexing and retrieval.
GMM uses K (basically 3 to 5) Gaussian models to represent the characteristics of each pixel point in the image, updates the Gaussian mixture model after a new frame of image is obtained, uses each pixel point in the current image to match with the Gaussian mixture model, and judges that the point is a background point if the matching is successful, otherwise, the point is a foreground point. Firstly, initializing several predefined Gaussian models, initializing parameters in the Gaussian models, and solving parameters to be used later. Secondly, processing each pixel in each frame to see whether the pixel is matched with a certain model or not, if so, classifying the pixel into the model, updating the model according to a new pixel value, and if not, establishing a Gaussian model by the pixel, initializing parameters and replacing the least probable model in the original model. And finally, selecting the front most possible models as background models to lay a cushion for background target extraction.
Fig. 2 is a schematic flow chart of a vehicle head and tail identification algorithm of a gaussian mixture model according to an embodiment of the present invention, as shown in fig. 2, including:
and 26, outputting the specific position of the vehicle head and the specific position of the vehicle tail.
The algorithm does not need to preset a head and tail template base and identify train types, but directly compares the currently acquired images of the train set with video frames for operation, so background modeling is needed firstly, namely, the background modeling is carried out when no image is acquired; and then, carrying out background difference to judge whether a foreground image appears or not, and positioning the specific position of the foreground image when the foreground image appears so as to determine the specific position of the head/tail of the vehicle.
Firstly, the acquired image data without any train information is used as a background image, and modeling is performed based on the GMM to be used as the background image.
And then, respectively starting from the first or last acquired train image, carrying out background difference, searching whether a foreground image appears, if so, continuously positioning the position of the foreground image, and outputting the specific positions of the train head and the train tail.
The train set fault identification method without the preset template provided by the embodiment of the invention utilizes the currently acquired current train image to identify the train head and the train tail by adopting a Gaussian mixture model without any preset train head and tail template library or train type identification, then utilizes the currently acquired full train image to identify the carriage connection part of the train set according to the correlation between the images and the characteristics of the image per se, and then carries out template matching with other carriage images of the train set to be detected, realizes fault identification on the premise of not searching a historical train template library according to the train type, completes real-time analysis and automatic early warning of the visual structure abnormal condition of the running motor train set, utilizes the currently acquired image of the same train to carry out fault identification, and not only avoids the influence of external interference factors such as illumination, driving speed and the like to a great extent, the method can not be influenced by manual repair modes such as position adjustment of parts and the like, effectively solves the problems of high false alarm rate, high missing report rate and the like caused by external factors such as illumination, train running speed and the like in the current fault identification, improves the efficiency of fault identification and maintenance operation of the motor train unit, and improves the discovery and early warning capability of hidden faults of the motor train unit in the running process.
On the basis of the embodiment, the carriage joints of the train set are obtained by detecting image symmetry recognition according to a carriage joint recognition algorithm.
Judging a carriage junction of each target image of the complete train set image from the image appearing on the locomotive;
for any target image in the complete train set images, if the fact that any target image comprises a carriage junction is judged and known, the left part of the carriage junction of any target image is used as a carriage of the train for splicing, and the right part of the carriage junction of any target image is used as the starting point of the carriage of the next train;
and if not, splicing any target image as a complete image.
The judgment of the carriage junction of each target image of the complete train set image specifically comprises the following steps:
for each target image, judging whether a vertical black area exists in the target image, and whether the range of the vertical black area is within a first preset threshold range;
for a target image which is provided with the vertical black area and the vertical black area is within the first preset threshold range, judging whether images of the target image on the left side and the right side of the vertical black area are mutually symmetrical;
and if the images of the left side and the right side of the vertical black area of the target image are mutually symmetrical through judgment, the vertical black area of the target image is a carriage joint.
Fig. 3 is a schematic flow chart of detecting image symmetry by a car junction identification algorithm according to an embodiment of the present invention, as shown in fig. 3, including:
and step 39, outputting the fused train image as a train compartment.
Firstly, judging the carriage joints of each target image based on the acquired images of the train head positions, splicing the left half section as the carriage joint if any target image comprises the carriage joint, splicing the right half section as the starting point of the next carriage, and splicing the whole image if the left half section is not the starting point of the next carriage.
According to the method and the device for recognizing the train compartment connection, the train compartment connection is recognized according to the characteristics of the train compartment based on the currently acquired images of the train set to be detected without any historical image template or train type recognition. The specific judgment method of the carriage junction is to judge whether any target image has a vertical black area, and the range of the vertical black area is within a first preset threshold range. If the target image exists, judging whether the images on the left side and the right side of the vertical black area are in mirror symmetry, if so, the black area of the target image is a carriage joint, otherwise, the target image does not include the carriage joint. The reason for this is that the train set has certain characteristics, and the identification of the carriage connection can be realized according to the characteristics of the train set. Taking a motor train unit as an example, the motor train unit is composed of at least two carriages with driving force and a plurality of carriages without traction force, wherein the two carriages with the driving force are two power carriages which are reversely symmetrical, namely a head and a tail of the motor train unit, the plurality of carriages without the traction force are carriages in the middle of the head and the tail, and the carriages in the middle are symmetrical to each other. The middle of any two carriages is the carriage joint. According to the acquired train group images, color images of carriages, carriage joints and backgrounds of the motor train unit are different, the carriages of the motor train unit can be firstly identified on the train group images, for other areas, whether the images on two sides of the area are symmetrical or not can be seen in order to distinguish the carriage joints from the background images, if the images are symmetrical, the area is the carriage joints, and if not, the area is the background image. And splicing the image with the spliced train image to form the carriage image.
The train set fault identification method without the preset template provided by the embodiment of the invention does not need any preset train head and tail template library and train type identification, but utilizes the currently acquired train image to firstly identify the train head and tail of the train set, then identifies the carriage joint of the train set according to the correlation between the images and the characteristics of the images, and then carries out template matching with other carriage images of the train set to be detected, realizes fault identification on the premise of not searching the historical train template library according to the train type, utilizes the currently acquired full train image to complete real-time analysis and automatic early warning of the visual structure abnormal condition of the running motor train set according to the correlation between the images and the characteristics of the images, utilizes the currently acquired image of the same train to carry out fault identification, not only avoids the influence of external interference factors such as illumination, driving speed and the like to a great extent, the method can not be influenced by manual repair modes such as position adjustment of parts and the like, effectively solves the problems of high false alarm rate, high missing report rate and the like caused by external factors such as illumination, train running speed and the like in the current fault identification, improves the efficiency of fault identification and maintenance operation of the motor train unit, and improves the discovery and early warning capability of hidden faults of the motor train unit in the running process.
On the basis of the above embodiment, before the feature matching of the target car image and the reference car image, the method further includes:
extracting Scale Invariant Feature Transform (SIFT) features of the target compartment image to obtain preset fault points;
and determining a preliminary fault area in the target compartment image according to the area position of the preset fault point.
And determining a preliminary fault area in the target compartment image according to the area position of the preset fault point, wherein the preliminary fault area is obtained through a non-maximum suppression algorithm.
Scale-Invariant Feature Transform (SIFT) is a description used in the field of image processing. The description has scale invariance, can detect key points in the image and is a local feature descriptor.
The SIFT feature detection mainly comprises the following 4 basic steps:
1. and (4) detecting an extreme value in a scale space, and searching image positions on all scales. Potential scale-and rotation-invariant points of interest are identified by gaussian derivative functions.
2. And (4) positioning the key points, and determining the position and the scale by fitting a fine model at each candidate position. The selection of the key points depends on their degree of stability.
3. And determining the direction, wherein one or more directions are allocated to each key point position based on the local gradient direction of the image. All subsequent operations on the image data are transformed with respect to the orientation, scale and location of the keypoints, providing invariance to these transformations.
4. Keypoint description local gradients of an image are measured at a selected scale in a neighborhood around each keypoint.
Non-Maximum Suppression (hereinafter abbreviated as NMS) is to suppress elements that are not Maximum as the name implies, and can be understood as a local Maximum search. The local representation is a neighborhood, and the neighborhood has two variable parameters, namely the dimension of the neighborhood and the size of the neighborhood. For example, in object detection, a sliding window is subjected to feature extraction, and after classification and identification by a classifier, each window obtains a classification and a score. But sliding windows can result in many windows containing or mostly crossing other windows. The NMS is then used to select the window with the highest score (the highest probability of being an object of a certain class) in the neighborhood and suppress those windows with low scores.
NMS non-maxima suppression is widely used in object detection, where the goal is to find the best location for object detection in order to eliminate redundant boxes.
The method for carrying out feature matching on the target carriage image and the reference carriage image and identifying the fault position of the target carriage specifically comprises the following steps:
according to the position of the preliminary fault area in the target compartment image, searching a corresponding position in the reference compartment image, and performing feature matching with the preliminary fault area to obtain a matching threshold value;
and if the matching threshold is judged to be smaller than a second preset threshold, the preliminary fault area contains a fault area, and the position of the fault area is output.
Fig. 4 is a schematic flow chart of a train set fault identification method without a preset template according to another embodiment of the present invention, as shown in fig. 4, including:
and step 410, finishing the fault judgment and outputting the position of the fault area.
Aiming at the currently collected complete train set image, based on the positions of interest points such as preset fault points and the like, preliminarily positioning a region which is possibly faulted in the image; and then, taking other image data of the train, which is not the carriage of the train, as an image template library, searching corresponding positions of the fault areas in the image template library, if the matching is successful, indicating that the image is a normal area, and if the matching threshold is low, indicating that the image contains the fault areas, and positioning the fault areas. The method specifically comprises the following steps:
firstly, extracting scale invariant feature transform descriptor (SIFT) features of a currently acquired train operation image;
then, a non-maximum suppression value (NMS) is adopted to position a preliminary fault area in the image according to the area position and the area fraction of interest points such as preset fault points and the like in the image;
and finally, matching and comparing the primary fault area with different carriages of the current train image, so as to judge whether the image really has the fault area and position the position of the fault area.
The train set fault identification method without the preset template provided by the embodiment of the invention does not need any historical template image, does not need to identify the train type, but adopts SIFT and NMS to preliminarily locate the fault area according to the current collected whole train image, further determines the specific position of the fault, completes the real-time analysis and automatic early warning of the visual structure abnormal condition of the running motor train unit according to the correlation between the images and the characteristics of the images, utilizes the current collected image of the same train to carry out fault identification, not only avoids the influence of external interference factors such as illumination, driving speed and the like to a great extent, but also can not be influenced by the manual repair mode such as position adjustment of parts and the like, effectively solves the problems of higher false alarm rate, higher missing report rate and the like caused by the external factors such as illumination, train running speed and the like in the current fault identification, improves the fault identification and the repair operation efficiency of the motor train unit, the hidden fault discovery and early warning capability of the motor train unit in the operation process is improved.
Fig. 5 is a schematic structural diagram of a train set fault identification apparatus without a preset template, as shown in fig. 5, including a first processing module 51, a second processing module 52, a splicing module 53 and an identification module 54, where:
the first processing module 51 is configured to identify a head and a tail of the train group according to the acquired train group image, and remove an image before the head and an image after the tail in the acquired train group image to obtain a complete train group image;
the second processing module 52 is configured to identify a car junction of the train group according to the complete train group image;
the splicing module 53 is used for splicing the images of the complete train set according to the carriage joints to obtain a complete image of each carriage of the train set;
the identification module 54 is configured to perform feature matching on the target car image and a reference car image, and identify a fault location of the target car, where the target car is a car to be detected, and the reference car is another car in the train set except the target car.
The existing train set fault identification method is based on a large number of train set and tail preset template libraries, images of various different types of train sets need to be collected before fault identification, the train sets and the tails of the train sets are detected according to the types of the train sets to be detected during fault identification, and corresponding historical templates are searched according to carriage numbers of the train sets, so that fault positions are further determined; and identifying the carriage joints of the train group according to the modified complete train group image, and splicing the collected images according to the positions of the carriage joints, so that the collected images are spliced and divided according to the carriage serial numbers to form a complete image of each carriage.
The specific operation is that for a group of train groups to be detected, when the train group is detected to appear, images of the train group are collected, the first processing module 51 takes the collected images as each frame of a video image, processes the images from a first image and a last image respectively, finds out images where the head appears and images where the tail appears, then judges the specific position of the head in the images where the head appears and the specific position of the tail in the images where the tail appears, and extracts the specific areas of the head and the tail. After the position of the head of the train and the position of the tail of the train are determined, images in front of the head of the train and images behind the tail of the train in the images of the train group are removed, and the purpose of the step is to remove the images which are irrelevant to the train in the images of the train group to obtain the images of the whole train group.
After the train head and the train tail of the train set are identified, the carriage connection part of the train set needs to be further identified. The train set has certain characteristics, and the second processing module 52 can realize the identification of the compartment connection part according to the characteristics of the train set. Taking a motor train unit as an example, the motor train unit is composed of at least two carriages with driving force and a plurality of carriages without traction force, wherein the two carriages with the driving force are two power carriages which are reversely symmetrical, namely a head and a tail of the motor train unit, the plurality of carriages without the traction force are carriages in the middle of the head and the tail, and the carriages in the middle are symmetrical to each other. The middle of any two carriages is the carriage joint. In the previous step, the image part irrelevant to the train in the train group image is removed to obtain the complete train group image. The color images of the carriages, the carriage joints and the background of the motor train unit are different, the carriages of the motor train unit can be firstly identified on the image of the complete train set, for other areas, whether the images on two sides of the area are symmetrical or not can be seen in order to distinguish the carriage joints from the background image, if the images are symmetrical, the area is the carriage joints, and if not, the area is the background image. The stitching module 53 identifies the car junction of each image and then stitches the images. If the image has no carriage joint, the stitching module 53 directly stitches the image, and if the image has a carriage joint, the stitching module 53 stitches the left half section of the image as the train carriage and stitches the right half section as the starting point of the next train carriage, so that the stitching of each train carriage of the motor train unit is completed.
After the train set head and tail and the carriage junction of the train set are identified, fault identification is needed, and the fault identification in the embodiment of the invention still adopts the train set image acquired on site instead of a preset template. Specifically, the identification module 54 takes out the corresponding stitched image of a certain car with a possible failure as a target car image, takes the image corresponding to the other car of the train group as a reference car image, performs feature matching on the target car image and the reference car image to obtain a matching value, and compares the matching value with a preset threshold value. If the matching value is higher than the preset threshold value, the target compartment is indicated to have no fault; otherwise, the target compartment is indicated to be in fault, and the fault position is output according to the matching result.
The apparatus provided in the embodiment of the present invention is used for executing the above method embodiments, and for detailed descriptions and specific processes, reference is made to the above method embodiments, which are not described herein again.
According to the train set fault recognition device without the preset template, provided by the embodiment of the invention, a train head and tail template library is not required to be preset, and the train type is not required to be recognized, but the currently collected train image is utilized, the train head and tail of the train set are firstly recognized, then the carriage connection part of the train set is recognized according to the correlation between the images and the characteristics of the images, and then the carriage connection part is subjected to template matching with other carriage images of the train set to be detected, so that fault recognition is realized on the premise of not searching a historical train template library according to the train type, and the real-time analysis and automatic early warning of the visual structure abnormal condition of the running motor train unit are completed. The method has the advantages that the current collected image of the same train is utilized to carry out fault recognition, so that the influence of external interference factors such as illumination, driving speed and the like is avoided to a great extent, the influence of manual repair modes such as position adjustment of parts and the like is avoided, the problems of high false alarm rate, high missing report rate and the like caused by external factors such as illumination, train running speed and the like in the current fault recognition are effectively solved, the efficiency of fault recognition and maintenance operation of the motor train unit is improved, and the discovery and early warning capability of hidden faults of the motor train unit in the running process is improved.
Fig. 6 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 6: a processor (processor)610, a communication Interface (Communications Interface)620, a memory (memory)630 and a bus 640, wherein the processor 610, the communication Interface 620 and the memory 630 complete communication with each other through the bus 640. Bus 640 may be used for information transfer between the electronic device and the sensor. The processor 610 may call logic instructions in the memory 630 to perform the following method: identifying the head and the tail of the train set according to the acquired images of the train set, and removing images in front of the head and images behind the tail in the acquired images of the train set to obtain images of the complete train set; identifying a carriage junction of the train group according to the complete train group image; splicing the images of the complete train set according to the carriage connection part to obtain a complete image of each carriage of the train set; and performing characteristic matching on the target compartment image and a reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
In addition, the logic instructions in the memory 630 may be implemented in software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
An embodiment of the present invention provides a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable a computer to execute the train set fault identification method without a preset template, which is provided in the foregoing embodiment, and for example, the method includes: identifying the head and the tail of the train set according to the acquired images of the train set, and removing images in front of the head and images behind the tail in the acquired images of the train set to obtain images of the complete train set; identifying a carriage junction of the train group according to the complete train group image; splicing the images of the complete train set according to the carriage connection part to obtain a complete image of each carriage of the train set; and performing characteristic matching on the target compartment image and a reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Various modifications and additions may be made to the described embodiments by those skilled in the art without departing from the spirit of the invention or exceeding the scope as defined in the appended claims.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A train set fault identification method without a preset template is characterized by comprising the following steps:
identifying the head and the tail of the train set according to the acquired images of the train set, and removing images in front of the head and images behind the tail in the acquired images of the train set to obtain images of the complete train set;
identifying a carriage junction of the train group according to the complete train group image;
splicing the images of the complete train set according to the carriage connection part to obtain a complete image of each carriage of the train set;
and performing characteristic matching on the target compartment image and a reference compartment image to identify the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
2. The method according to claim 1, wherein the locomotive and the tail of the train consist are identified according to a locomotive and tail identification algorithm of a Gaussian mixture model, and the method specifically comprises the following steps:
based on a Gaussian mixture model, learning and modeling are carried out on image data without train information to obtain a background image;
performing background difference on the train set image by combining the background image to obtain an image of the train head and an image of the train tail;
and determining the specific position of the vehicle head in the image where the vehicle head appears, extracting a vehicle head image area, determining the specific position of the vehicle tail in the image where the vehicle tail appears, and extracting a vehicle tail image area.
3. The method according to claim 2, wherein the car junctions of the train consist are obtained by detecting image symmetry according to a car junction identification algorithm, and specifically comprises:
judging a carriage junction of each target image of the complete train set image from the image appearing on the locomotive;
for any target image in the complete train set images, if the fact that any target image comprises a carriage junction is judged and known, the left part of the carriage junction of any target image is used as a carriage of the train for splicing, and the right part of the carriage junction of any target image is used as the starting point of the carriage of the next train;
and if not, splicing any target image as a complete image.
4. The method according to claim 3, wherein the determining of the car junction for each target image of the complete train set image specifically comprises:
for each target image, judging whether a vertical black area exists in the target image, and whether the range of the vertical black area is within a first preset threshold range;
for a target image which is provided with the vertical black area and the vertical black area is within the first preset threshold range, judging whether images of the target image on the left side and the right side of the vertical black area are mutually symmetrical;
and if the images of the left side and the right side of the vertical black area of the target image are mutually symmetrical through judgment, the vertical black area of the target image is a carriage joint.
5. The method according to claim 1, wherein prior to said feature matching the target car image with the reference car image, the method further comprises:
extracting Scale Invariant Feature Transform (SIFT) features of the target compartment image to obtain preset fault points;
and determining a preliminary fault area in the target compartment image according to the area position of the preset fault point.
6. The method according to claim 5, wherein the determining of the preliminary fault area in the target car image according to the area location of the preset fault point is performed by a non-maximum suppression algorithm.
7. The method according to claim 5 or 6, wherein the step of performing feature matching on the target car image and the reference car image to identify the fault location of the target car comprises the following specific steps:
according to the position of the preliminary fault area in the target compartment image, searching a corresponding position in the reference compartment image, and performing feature matching with the preliminary fault area to obtain a matching threshold value;
and if the matching threshold is judged to be smaller than a second preset threshold, the preliminary fault area contains a fault area, and the position of the fault area is output.
8. The utility model provides a train set fault recognition device that need not to predetermine template which characterized in that includes:
the first processing module is used for identifying the head and the tail of the train set according to the collected images of the train set, and removing images in front of the head and images behind the tail in the collected images of the train set to obtain images of the whole train set;
the second processing module is used for identifying the carriage connection part of the train set according to the complete train set image;
the splicing module is used for splicing the images of the complete train set according to the carriage joints to obtain a complete image of each carriage of the train set;
and the identification module is used for carrying out feature matching on the target compartment image and the reference compartment image and identifying the fault position of the target compartment, wherein the target compartment is a compartment to be detected, and the reference compartment is other compartments except the target compartment in the train set.
9. An electronic device, comprising a memory and a processor, wherein the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 7.
10. A non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the train consist fault identification method according to any one of claims 1 to 7 without a preset template.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811320495.0A CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811320495.0A CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740410A CN109740410A (en) | 2019-05-10 |
| CN109740410B true CN109740410B (en) | 2020-11-13 |
Family
ID=66355666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811320495.0A Active CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740410B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110276258A (en) * | 2019-05-21 | 2019-09-24 | 智慧互通科技有限公司 | A kind of method and system identifying vehicle appearance |
| CN110285870A (en) * | 2019-07-22 | 2019-09-27 | 深圳市卓城科技有限公司 | Vehicle spindle-type and wheel number determination method and its system |
| CN111860303A (en) * | 2020-07-17 | 2020-10-30 | 成都铁安科技有限责任公司 | Train pantograph image acquisition system and control method thereof |
| CN111860384B (en) * | 2020-07-27 | 2021-12-14 | 上海福赛特智能科技有限公司 | Vehicle type recognition method |
| CN111862623A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Vehicle side map splicing device and method |
| CN112183786A (en) * | 2020-10-19 | 2021-01-05 | 邦邦汽车销售服务(北京)有限公司 | Method and device for determining damaged part of vehicle |
| CN113837160B (en) * | 2021-11-29 | 2022-04-22 | 天津市中环系统工程有限责任公司 | Method for identifying normally-living people and vehicles without preset information |
| CN115063903A (en) * | 2022-06-08 | 2022-09-16 | 深圳市永达电子信息股份有限公司 | Railway freight train compartment body abnormity monitoring method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0579133A1 (en) * | 1992-07-15 | 1994-01-19 | Hitachi, Ltd. | Superconducting magnet abnormality detection and protection apparatus |
| EP0800153A3 (en) * | 1996-04-01 | 1998-05-13 | Hamamatsu Photonics K.K. | Smoke detecting apparatus and method |
| CN104463235A (en) * | 2014-11-18 | 2015-03-25 | 中国铁道科学研究院电子计算技术研究所 | Fault recognition method and device based on operation images of motor train unit |
| CN105787925A (en) * | 2016-02-03 | 2016-07-20 | 中国科学院空间应用工程与技术中心 | Push-scan type optical remote sensing load original image bad line automatic detection method and system |
-
2018
- 2018-11-07 CN CN201811320495.0A patent/CN109740410B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0579133A1 (en) * | 1992-07-15 | 1994-01-19 | Hitachi, Ltd. | Superconducting magnet abnormality detection and protection apparatus |
| EP0800153A3 (en) * | 1996-04-01 | 1998-05-13 | Hamamatsu Photonics K.K. | Smoke detecting apparatus and method |
| CN104463235A (en) * | 2014-11-18 | 2015-03-25 | 中国铁道科学研究院电子计算技术研究所 | Fault recognition method and device based on operation images of motor train unit |
| CN105787925A (en) * | 2016-02-03 | 2016-07-20 | 中国科学院空间应用工程与技术中心 | Push-scan type optical remote sensing load original image bad line automatic detection method and system |
Non-Patent Citations (2)
| Title |
|---|
| A multi-camera image processing and visualization system for train safety assessment;Giuseppe Lisanti et al.;《arXiv》;20150731;全文 * |
| 动车组运行状态智能检测装备;张望等;《铁路计算机应用》;20180731;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740410A (en) | 2019-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109740410B (en) | Train set fault identification method and device without preset template | |
| CN111488875B (en) | Vehicle insurance claim settlement loss checking method and device based on image recognition and electronic equipment | |
| CN107729818B (en) | Multi-feature fusion vehicle re-identification method based on deep learning | |
| CN104463235B (en) | Fault recognition method and device based on EMU operation image | |
| TWI430212B (en) | Abnormal behavior detection system and method using automatic classification of multiple features | |
| CN103699905B (en) | Method and device for positioning license plate | |
| CN110119701B (en) | Visual relationship detection-based coal mine fully mechanized coal mining face unsafe behavior identification method | |
| KR20190063839A (en) | Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process | |
| CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
| KR101653278B1 (en) | Face tracking system using colar-based face detection method | |
| Wang et al. | An effective method for plate number recognition | |
| CN104615986A (en) | Method for utilizing multiple detectors to conduct pedestrian detection on video images of scene change | |
| CN112364740B (en) | Unmanned aerial vehicle room monitoring method and system based on computer vision | |
| Nguyen et al. | Vehicle re-identification with learned representation and spatial verification and abnormality detection with multi-adaptive vehicle detectors for traffic video analysis. | |
| JP5931662B2 (en) | Road condition monitoring apparatus and road condition monitoring method | |
| CN110210433B (en) | Container number detection and identification method based on deep learning | |
| Chen et al. | Automatic detection of traffic lights using support vector machine | |
| CN107436906A (en) | A kind of information detecting method and device | |
| CN109543498B (en) | Lane line detection method based on multitask network | |
| Mijić et al. | Traffic sign detection using yolov3 | |
| CN111325265A (en) | Detection method and device for tampered image | |
| CN115797970B (en) | Dense pedestrian target detection method and system based on YOLOv5 model | |
| CN113657378B (en) | Vehicle tracking method, vehicle tracking system and computing device | |
| Eleuch et al. | Feature-based vehicle tracking at roundabouts in visual sensor networks | |
| CN114937248A (en) | Vehicle tracking method and device for cross-camera, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |