CN116580245A - Rail wagon bearing saddle dislocation fault identification method - Google Patents
Rail wagon bearing saddle dislocation fault identification method Download PDFInfo
- Publication number
- CN116580245A CN116580245A CN202310615378.1A CN202310615378A CN116580245A CN 116580245 A CN116580245 A CN 116580245A CN 202310615378 A CN202310615378 A CN 202310615378A CN 116580245 A CN116580245 A CN 116580245A
- Authority
- CN
- China
- Prior art keywords
- key point
- point group
- image
- bearing saddle
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000001514 detection method Methods 0.000 claims abstract description 71
- 230000002708 enhancing effect Effects 0.000 claims abstract description 5
- 238000006073 displacement reaction Methods 0.000 claims abstract description 4
- 238000003379 elimination reaction Methods 0.000 claims description 49
- 230000008030 elimination Effects 0.000 claims description 31
- 238000012549 training Methods 0.000 claims description 15
- 238000001914 filtration Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 230000003321 amplification Effects 0.000 claims description 6
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 235000017274 Diospyros sandwicensis Nutrition 0.000 claims description 5
- 241000282838 Lama Species 0.000 claims description 5
- 230000000903 blocking effect Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 2
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 8
- 238000000605 extraction Methods 0.000 description 7
- 238000007689 inspection Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000007637 random forest analysis Methods 0.000 description 3
- 230000008439 repair process Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Biodiversity & Conservation Biology (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method for identifying dislocation faults of a bearing saddle of a railway wagon, which solves the problem that the existing deep learning method is low in accuracy in detecting key points of the bearing saddle and belongs to the field of railway wagon fault detection. The invention comprises the following steps: acquiring an image to be detected, intercepting a rough positioning image of a bearing saddle in the image to be detected, and enhancing contrast, corner points and boundary characteristics of the rough positioning image; and obtaining key point groups of 8 key points of the bearing saddle, namely an upper left key point, an upper right key point, a lower left key point, a lower right key point, a left upper flange and an upper right flange on the enhanced image by using a key point detection model, calculating the inclination angle of the bearing saddle by using the 8 key points, and judging whether dislocation faults occur according to whether the inclination angle is larger than or equal to a threshold value. If the displacement occurs, fault alarm is carried out, otherwise, the bearing saddle of the image to be tested is not displaced.
Description
Technical Field
The invention relates to a method for identifying dislocation faults of a bearing saddle of a railway wagon, and belongs to the technical field of railway wagon fault detection.
Background
The bearing adapter is a support for the vehicle wheel pair bearing assembly for transferring loads from the vehicle body and plays a very important role in the vehicle truck. If the bearing saddle is misplaced, the bearing saddle and the front cover of the outer rolling bearing are seriously worn, so that serious vehicle fault hidden danger is caused, and the driving safety is endangered. In the bearing saddle dislocation fault detection, a manual image inspection mode is adopted for fault detection. Considering that the horizontal inclination of the bearing saddle is larger than or equal to an angle threshold value, namely the bearing saddle is judged to be misplaced, the fault detection difficulty is increased by a smaller angle threshold value (such as 3 degrees); poor image quality such as image blurring, poor contrast ratio, camera angle difference and the like existing in different detection stations causes the problem of non-ideal recognition effect; in addition, the conditions such as fatigue and omission are very easy to occur in the working process of the car inspection personnel, so that the occurrence of missed inspection and false inspection is caused, and the driving safety is influenced. In recent years, deep learning and artificial intelligence are continuously developed, and the technology is continuously mature. Therefore, the carrying saddle dislocation fault identification is carried out by deep learning, and the detection efficiency and the accuracy can be effectively improved. However, the existing deep learning method has the problem that the detection accuracy of key points of the bearing saddle is not high.
Disclosure of Invention
Aiming at the problem that the key point detection accuracy of the conventional deep learning method on the bearing saddle is not high, the invention provides a method for identifying dislocation faults of the bearing saddle of a railway wagon
The invention discloses a method for identifying dislocation faults of a bearing saddle of a railway wagon, which comprises the following steps:
s1, acquiring an image to be detected of a vehicle passing, intercepting a rough positioning image of a bearing saddle in the image of the vehicle passing, and enhancing contrast, corner points and boundary characteristics of the rough positioning image;
s2, obtaining key point groups of 8 key points of the bearing saddle, namely an upper left, an upper right, a lower left, a lower right, a lower left intersected with a bearing, a lower right intersected with the bearing, an upper left and an upper right on a flange, on the enhanced bearing saddle image by using a key point detection model, judging whether the number of the key points in the current key point group is 8, if so, calculating the inclination angle of the bearing saddle by using the 8 key points, and turning to S3; if not, carrying out fault alarm;
the key point detection model comprises a contour detection model, a key point estimation network and a redundancy elimination model;
the contour detection model extracts the image contour of the enhanced bearing saddle image, a key point estimation network adds bearing saddle component frames to the extracted image contour to obtain bearing saddle component key points, if 1 bearing saddle component frame is obtained, the bearing saddle component frames are subjected to key point estimation to obtain key point groups of 8 key points, and the key point groups of 8 key points are final key point groups and are used for calculating the inclination angle of the bearing saddle; if a plurality of bearing saddle component frames are obtained, carrying out key point estimation on each bearing saddle component frame to obtain 8 key point groups of key points corresponding to each bearing saddle component frame, and carrying out confidence similarity elimination, edge relative distance similarity elimination and/or position similarity elimination on the obtained key point groups by a redundancy elimination model to obtain a final key point group for calculating the inclination angle of the bearing saddle;
s3, judging whether dislocation faults occur according to whether the inclination angle is larger than or equal to a threshold value; if the displacement occurs, fault alarm is carried out, otherwise, the bearing saddle of the image to be detected is not displaced.
Preferably, the key point estimation network is implemented by adopting a CNN network based on SPPE.
Preferably, the confidence eliminating method comprises the following steps:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the confidence degree similarity exceeds a set threshold, the corresponding target key point group P j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The confidence similarity of (2) is calculated by:
if the key point group P j Coordinates of the nth key point in (a)At->Within the range, target key point group P j N-th key point of the key point group P i The confidence similarity of the corresponding nth key point is +.>Otherwise, 0, counting the target key point group P j 8 key points and key point group P i The sum of confidence similarity of the corresponding 8 key points is taken as a key point group P i And target key point group P j Confidence similarity of (c);
σ 1 representing the confidence redundancy elimination parameter,and->Respectively represent key point groups P i And target key point group P j Confidence of the nth key point, < ->Representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
Preferably, the method for eliminating the relative distance of the edge comprises the following steps:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the relative distance of the target key point group P exceeds the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The relative distance of (2) is calculated by:
if the key point group P j Coordinates of the nth key point in (a)At->Within and->Distance from nearest contour is greater than +.>Target key point group P j N-th key point of the key point group P i The relative distance of the corresponding nth key point is 1, n=1, 2, …,8, otherwise is 0, and the target key point group P is counted j 8 key points and key point group P i The sum of the relative distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Relative distance of (2);
Representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
Preferably, the method for eliminating the key points close to each other comprises the following steps:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the spatial distance of the target key point group P is not more than the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The spatial distance of (2) is calculated by:
target key point group P j N-th key point of the key point group P i The spatial distance of the corresponding nth key point isStatistics of target keypoint group P j 8 key points and key point group P i The sum of the spatial distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Is a spatial distance of (2);
σ 2 for the spatial distance redundancy elimination parameter,representing a set of keypoints P i Coordinates of the nth key point.
Preferably, the contour detection model comprises a blocking unit, a gradient lifting tree classifier and a synthesis unit, and the contour detection model is used for obtaining the contour of the bearing saddle in the image, and the method comprises the following steps:
the blocking unit blocks the reinforced bearing saddle image by adopting a sliding window to obtain a plurality of image blocks, the image blocks are input into a gradient lifting tree classifier, the gradient lifting tree classifier outputs labels of all the image blocks, the labels of all the image blocks are input into a comprehensive unit, and the comprehensive unit obtains the contour of the bearing saddle in the image according to the labels.
Preferably, training the contour detection model includes:
the method comprises the steps of partitioning an enhanced bearing saddle image by adopting a sliding window, obtaining the characteristics of each image block, calculating the label of each image block by adopting a PCA algorithm, taking the characteristics of the image block as input and the label as output, and forming a training set to train the gradient lifting tree classifier.
Preferably, the method of obtaining the feature of each image block includes:
processing the image block to enable each pixel to have K-dimensional characteristics, wherein the K-dimensional characteristics comprise color channel characteristics, gradient amplitude channel characteristics, high-pass filtering channel characteristics, maximum fuzzy filtering channel characteristics and Gabor filtering channel characteristics with different dimensions in multiple directions;
and performing dimension reduction on the image block characteristics to obtain dimension-reduced image block characteristics.
Preferably, the method further comprises establishing a keypoint detection sample data set, training a keypoint detection model using the keypoint detection sample data set, wherein the method for removing the red frame in the bearing saddle image of the oblique fault when establishing the keypoint detection sample data set comprises:
setting a threshold Th, extracting the region where the red frame is located, and obtaining a white Mask image after large-core morphological expansion;
and repairing a white area in the white Mask image by using an image repairing model lama so as to remove a red frame in the inclined fault bearing saddle image.
Preferably, the critical point detection sample data set is a data set after sample amplification, and the amplification method comprises the following steps: sharpening, contrast enhancement, rotation, translation, scaling, and mirroring of images.
The invention has the beneficial effects that the fault detection mode based on deep learning key point detection is adopted to replace manual detection of the dislocation of the bearing saddle, so that the detection efficiency and the accuracy are improved. The invention can improve the accuracy and precision of the detection of the bearing saddle key points, reduce the difference of camera angles of different detection stations and effectively improve the accuracy of the system based on the application of the key point detection model with the bearing saddle component frame to the detection of the bearing saddle key points. Aiming at the problem of unsatisfactory recognition effect caused by poor image quality such as image blurring and poor contrast ratio of different detection stations, the system adopts the method of firstly enhancing and then detecting key points, and has high robustness. According to the invention, the training set is added after the true fault image with the red frame is removed by a deep learning mode, so that the problem of serious unbalance of positive and negative samples is solved, the approximation degree of the training set to the data distribution of the bearing saddle is enhanced, and the fault recognition rate can be effectively improved.
Drawings
FIG. 1 is a flow chart of fault identification in accordance with the present invention;
fig. 2 shows an example of an original image of a bearing saddle and a key point mark.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other. The invention is further described below with reference to the drawings and specific examples, which are not intended to be limiting.
The method for identifying the dislocation faults of the bearing saddle of the railway wagon comprises the steps of establishing a key point detection model, detecting key point groups of 8 key points of an upper left, an upper right, an upper left, a lower right, a lower left, a lower right, a left upper flange and a right upper flange of a bearing on an image of the bearing saddle by using the key point detection model, and determining whether the dislocation faults are caused by the fact that the number of the key points in the key point groups is 8, and calculating the inclination angles of the bearing saddle according to the 8 key points, otherwise, giving fault alarm;
the method of the embodiment specifically comprises the following steps:
step 1, establishing a key point detection model data set
High-definition equipment is built around the train track respectively, and high-definition images are acquired after a truck passes through the equipment. The image is a clear truck bogie side gray scale image. The form of the saddle bearing parts varies among different types of bogies. Therefore, the bearing saddle component structures of all types of bogies such as K2, K6, K4, K5, Z8AB and the like are summarized so that the training data distribution approximates to all bearing saddle data distribution as much as possible, and the establishment of a perfect key point detection sample data set is a necessary premise for constructing a high-precision key point detection model. The normal fault-free bearing saddle image can be simply obtained, the data set of the inclined fault bearing saddle is few, and the positive and negative samples are seriously unbalanced.
Interference information such as red frames in a true fault image of the inclined fault bearing saddle cannot be directly used as an image training set. To solve this problem, the red frame in the true failure image needs to be removed. For a long time, a large number of researchers have been researching how to better remove elements in a picture and correctly replace the background, and this task is also called image restoration, in which a deep learning mode is adopted to remove a red frame. According to the method, based on the obvious gray value characteristic of the image to be repaired in the red channel of the red frame, the Mask image with the white region of the red frame can be approximately extracted by setting the threshold Th; the white Mask map is expanded by large kernel morphology to obtain a training Mask to be repaired by the final lama, and the white area in the white Mask map is repaired by using the image repair model lama so as to remove the red frame in the oblique fault bearing saddle image. The main innovation points of the lama are as follows: a new repair network structure is provided, and fast Fourier convolution is used, so that the repair network structure has the advantages of wide image receiving domain, high receptive field perception loss and large training mask (mask), and the performance potential of the first two components can be effectively improved. It can be well generalized to higher resolution images than during training, achieving comparable performance to the baseline with lower parameter amounts and computational costs.
As the truck components can be affected by natural or artificial conditions such as rainwater, mud, oil stain, black paint and the like; there may be differences in the images taken at different sites. Thus, there are a great deal of differences between the bearing saddle images. Therefore, in the process of collecting the saddle image data, the saddle images under various conditions are collected as much as possible while ensuring the diversity. The keypoint detection model sample dataset comprises: a gray image set and a mark image set. The gray image set is a high ash removal image shot by the equipment. The original image of the marked image set is a coarse positioning image of the bearing saddle, the marked data set is a json file, and the marked data set is obtained in a manual marking mode. The gray image data sets and the marking data sets are in one-to-one correspondence, namely each gray image corresponds to one marking file, and the json file is marked with accurate position information of 8 key points of the bearing saddle in detail.
The establishment of the sample data set of the key point detection model includes images under various conditions, but in order to improve the stability of the algorithm, data amplification needs to be performed on the sample data set. The amplification form comprises operations of sharpening, contrast enhancement, rotation, translation, scaling, mirroring and the like of the image, and each operation is carried out under random conditions, so that the diversity and applicability of the sample can be ensured to the greatest extent.
Step 2: establishing a key point detection model
In the embodiment, 8 key point detection methods of upper left, upper right, lower left, lower right, lower left intersecting with a bearing, lower right intersecting with a bearing, upper left of a flange and upper right of a flange of an adapter in an image are positioned by using a key point detection model. This approach is a top-down approach that detects saddle objects in the image and then detects boundary keypoints for each saddle region independently. The key point detection model of the embodiment comprises a contour detection model, a key point estimation network and a redundancy elimination model; firstly, extracting an image contour of an enhanced bearing saddle image by using a contour detection model, adding a bearing saddle part frame to the extracted image contour by using a key point estimation network to obtain bearing saddle part key points, and if 1 bearing saddle part frame is obtained, carrying out key point estimation on the bearing saddle part frame to obtain key point groups of 8 key points, wherein the key point groups of 8 key points are final key point groups and are used for calculating the inclination angle of a bearing saddle; if a plurality of bearing saddle component frames are obtained, carrying out key point estimation on each bearing saddle component frame to obtain 8 key point groups of key points corresponding to each bearing saddle component frame, and carrying out confidence similarity elimination, edge relative distance similarity elimination and/or position similarity elimination on the obtained key point groups by a redundancy elimination model to obtain a final key point group for calculating the inclination angle of the bearing saddle;
the bearing saddle key points used for calculating the angle are all positioned on the bearing saddle boundary, the brightness difference of images at two sides of the bearing saddle boundary is obvious compared with the brightness difference in the same component, and the detection efficiency and accuracy of the key points are improved through the two aspects of candidate frame generation and redundant key point elimination of the contour features in the target area.
In a preferred embodiment, the contour detection model in this embodiment includes a block unit, a gradient lifting tree classifier, and a synthesis unit, where the contour detection model is used to obtain a contour of a bearing saddle in an image, and the specific method includes:
the contour edge of the bearing saddle component is in a local structure form of straight line or T-shaped combination, and a better target area candidate frame can be obtained from an edge extraction image through an edge recombination strategy. The boundary extraction method in this embodiment adopts an edge extraction algorithm based on PCA and structured random forests. The problem of missing detection of partial weak edges in the connection of the high and low threshold values exists in the traditional contour boundary extraction, and the edge extraction algorithm based on PCA and the structured random forest can keep the weak edge points of the bearing saddle with good connectivity, so that the extracted image contour is more accurate and complete. The segmentation unit segments the reinforced bearing saddle image by adopting a sliding window to obtain the characteristics of a plurality of image blocks, the image contour is manually marked during training in the embodiment to obtain the characteristics of each image block, the PCA algorithm is adopted to calculate the label of each image block, the characteristics of the image block are taken as input and the label is taken as output to form a training set to train the gradient lifting tree classifier, the sliding window is adopted to send the image blocks into the trained gradient lifting tree classifier to obtain the labels of each image block during contour extraction, the labels of each image block are input into the comprehensive unit, and the comprehensive unit obtains the bearing saddle contour in the image according to the labels.
The features of each original image block (32 x 32) are calculated. Firstly, processing a certain original image block to enable each pixel to have K-dimensional information, wherein K is 29, the K corresponds to 1 color channel (gray level image), 2 gradient amplitude channels, 1 high-pass filtering channel, 1 maximum fuzzy filtering channel and 24 Gabor filtering channels (corresponding to 12 directions, each direction corresponds to 2 different scales), and then, dimension reduction is carried out on the image block characteristics of the whole original image block: 32×32×29/4=7424 dimensions. Plus each channel contains paired pixel information at 5 x 5 resolution, each image block thus contains 7424+29×=16124 dimension features altogether. To this end, each feature corresponds to an edge block and a label (0 or 1). The Gabor filter can well extract the edge information of the image, is insensitive to illumination change, and adopts Gabor with multiple scales and directions to carry out convolution processing on the image so as to simulate a multi-scale mechanism of processing the image by the human brain.
In a preferred embodiment, the key point estimation network of the present embodiment is implemented by using a CNN network based on SPPE.
The redundant boundary box generates redundant boundary points, and the redundancy elimination rule of the embodiment is composed of three parts of confidence elimination, edge relative distance elimination and distance elimination. The redundant set of keypoints is eliminated as long as one of the elimination criteria is met. Even if the positioning of the bearing saddle boundary frame is wrong, the correct estimation can be carried out, and the efficiency and the accuracy of the bearing saddle target detection in the embodiment are effectively improved.
In a preferred embodiment, the method for eliminating confidence coefficient of the present embodiment includes:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the confidence degree similarity exceeds a set threshold, the corresponding target key point group P j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The confidence similarity of (2) is calculated by:
if the key point group P j Coordinates of the nth key point in (a)At->Within the range, target key point group P j N-th key point of the key point group P i The confidence similarity of the corresponding nth key point is +.>Otherwise, 0, counting the target key point group P j 8 key points and key point group P i The sum of confidence similarity of the corresponding 8 key points is taken as a key point group P i And target key point group P j Confidence similarity of (c);
σ 1 representing the confidence redundancy elimination parameter,and->Respectively represent key point groups P i And target key point group P j Confidence of the nth key point, < ->Representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
The confidence coefficient eliminating principle is to count the total number of the confidence coefficient similarity of the key points in the two key point groups and eliminate the key point groups with the similar confidence coefficient.
Edge relative distance elimination, i.e., elimination of keypoints that are far from the nearest edge contour. The image profile is obtained by an edge extraction algorithm based on PCA and structured random forests. In a preferred embodiment, the method for eliminating the edge relative distance in this embodiment:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the relative distance of the target key point group P exceeds the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the above elimination process until the targetThe key point group is empty;
wherein, the key point group P i And target key point group P j The relative distance of (2) is calculated by:
a represents the outline of the bearing saddleThe nearest point; if the key point group P j Coordinates of the nth key point +.>At the position ofWithin and->Distance from nearest contour is greater than +.>Target key point group P j N-th key point of the key point group P i The relative distance of the corresponding nth key point is 1, n=1, 2, …,8, otherwise is 0, and the target key point group P is counted j 8 key points and key point group P i The sum of the relative distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Is a relative distance of (2);
representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
Edge relative distanceThe method for realizing the separation elimination is similar to the key point group with similar elimination confidence coefficient, and only calculatesComparison with each otherDistance from edge contour, when ∈>The function is 1 when the distance from the nearest edge is farther, otherwise, the function is 0; if the point far from the nearest edge exceeds 3, the target key point group P j Will be eliminated.
In a preferred embodiment, the method for eliminating the keypoints with similar positions in this embodiment is to calculate the sum of the spatial distances between two keypoint groups, and select the reference keypoint group P with higher confidence i If the target key point group P j And reference key point group P i Closer distance, which means higher overlap of the two, then the target set of keypoints P j Will be eliminated. The method specifically comprises the following steps:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the spatial distance of the target key point group P is not more than the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The spatial distance of (2) is calculated by:
target key point group P j N-th key point of the key point group P i The spatial distance of the corresponding nth key point isStatistics of target keypoint group P j 8 key points and key point group P i The sum of the spatial distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Is a spatial distance of (2);
σ 2 for the spatial distance redundancy elimination parameter,representing a set of keypoints P i Coordinates of the nth key point.
And step 3, judging whether dislocation faults occur according to whether the inclination angle is larger than or equal to a threshold value. If the displacement occurs, fault alarm is carried out, otherwise, the bearing saddle of the image to be detected is not displaced.
Firstly, a real passing image is intercepted into a coarse positioning image according to the prior, and then the coarse positioning image is subjected to image enhancement to obtain an enhanced image with obvious contrast and obvious angular points and boundary characteristics. And 8 key points of the bearing saddle, such as upper left, upper right, lower left, lower right, lower left intersected with the bearing, lower right intersected with the bearing, upper left baffle edge and upper right baffle edge, can be obtained by using the key point detection model on the enhanced image. When the number of the key points is incorrect, directly alarming; and under the condition that the number of the key points is correct, calculating the inclination angle of the current bearing saddle by using 8 key points, and judging whether dislocation faults occur according to whether the inclination angle is larger than or equal to a threshold value. If dislocation occurs, performing fault alarm; and if no fault exists, processing the next bearing saddle image.
In view of the high degree of fineness to be identified in the detection of the bearing saddle fault, for example, the user requires 3 degrees or more, namely, alarming, the component is difficult to identify. The problem of unsatisfactory recognition effect caused by poor image quality such as image blurring and poor contrast ratio existing in different detection stations is solved, and the system adopts the steps of enhancing and then detecting key points of components so as to improve the accuracy. The precision of the detection of the key points of the bearing saddle directly influences the calculation of the inclination angle of the part, and the final error fault recognition accuracy of the bearing saddle is related. The system first performs an average luminance analysis on the coarsely positioned image, correcting the image luminance histogram to standard luminance to enhance the image contrast. And then guiding and filtering the image with enhanced contrast to obtain a final enhanced image. And finally, the corner and edge features in the enhanced image are obviously easy to detect key points. The guide filtering can well keep the details of each edge area in the rough positioning image, and the areas inside and outside the edges are smoothed much.
The inclination angle of the bearing adapter is calculated as follows. First, the inclination angles angle1 to angle4 of 4 straight lines are calculated, wherein the 4 lines are respectively two points of upper left and upper right, lower left and lower right, intersecting with the bearing (lower left intersecting with the bearing, lower right intersecting with the bearing), upper left of the flange and upper right of the flange. And then calculating 4 angle averages to obtain the inclination angle of the whole bearing saddle.
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein may be combined in ways other than as described in the original claims. It is also to be understood that features described in connection with separate embodiments may be used in other described embodiments.
Claims (10)
1. A method for identifying a dislocation fault of a bearing saddle of a railway wagon, which is characterized by comprising the following steps:
s1, acquiring an image to be detected of a vehicle passing, intercepting a rough positioning image of a bearing saddle in the image of the vehicle passing, and enhancing contrast, corner points and boundary characteristics of the rough positioning image;
s2, obtaining key point groups of 8 key points of the bearing saddle, namely an upper left, an upper right, a lower left, a lower right, a lower left intersected with a bearing, a lower right intersected with the bearing, an upper left and an upper right on a flange, on the enhanced bearing saddle image by using a key point detection model, judging whether the number of the key points in the current key point group is 8, if so, calculating the inclination angle of the bearing saddle by using the 8 key points, and turning to S3; if not, carrying out fault alarm;
the key point detection model comprises a contour detection model, a key point estimation network and a redundancy elimination model;
the contour detection model extracts the image contour of the enhanced bearing saddle image, a key point estimation network adds bearing saddle component frames to the extracted image contour to obtain bearing saddle component key points, if 1 bearing saddle component frame is obtained, the bearing saddle component frames are subjected to key point estimation to obtain key point groups of 8 key points, and the key point groups of 8 key points are final key point groups and are used for calculating the inclination angle of the bearing saddle; if a plurality of bearing saddle component frames are obtained, carrying out key point estimation on each bearing saddle component frame to obtain 8 key point groups of key points corresponding to each bearing saddle component frame, and carrying out confidence similarity elimination, edge relative distance similarity elimination and/or position similarity elimination on the obtained key point groups by a redundancy elimination model to obtain a final key point group for calculating the inclination angle of the bearing saddle;
s3, judging whether dislocation faults occur according to whether the inclination angle is larger than or equal to a threshold value; if the displacement occurs, fault alarm is carried out, otherwise, the bearing saddle of the image to be tested is not displaced.
2. The method for identifying the dislocation faults of the bearing saddle of the railway wagon according to claim 1, wherein the key point estimation network is realized by adopting a CNN network based on SPPE.
3. The method for identifying dislocation faults of bearing saddles of railway wagons according to claim 1, wherein the method for eliminating confidence comprises the following steps:
ordering the confidence degrees of the key point groups, and selecting the key with the highest confidence degreePoint group as reference Point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the confidence degree similarity exceeds a set threshold, the corresponding target key point group P j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The confidence similarity of (2) is calculated by:
if the key point group P j Coordinates of the nth key point in (a)At->Within the range, target key point group P j N-th key point of the key point group P i The confidence similarity of the corresponding nth key point is +.>Otherwise, 0, counting the target key point group P j 8 key points and key point group P i The sum of confidence similarity of the corresponding 8 key points is taken as a key point group P i And target key point group P j Confidence similarity of (c);
σ 1 representing the confidence redundancy elimination parameter,and->Respectively represent key point groups P i And target key point group P j Confidence of the nth key point, < ->Representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
4. The method for identifying dislocation faults of bearing saddles of railway wagons according to claim 1, characterized by the method for eliminating the relative distance of edges:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the relative distance of the target key point group P exceeds the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the elimination process until the target key point group is empty;
wherein, the key point group P i And target key point group P j The relative distance of (2) is calculated by:
if the key point group P j Coordinates of the nth key point in (a)At->Within and->Distance from nearest contour is greater than +.>Target key point group P j N-th key point of the key point group P i The relative distance of the corresponding nth key point is 1, n=1, 2, …,8, otherwise is 0, and the target key point group P is counted j 8 key points and key point group P i The sum of the relative distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Is a relative distance of (2);
representing a set of keypoints P i Coordinates of the nth key point, +.>Representing a set of keypoints P i 1/10 of the detection frame.
5. The method for identifying the dislocation faults of the bearing saddle of the railway wagon according to claim 1, wherein the method for eliminating the key points which are close in position comprises the following steps:
ordering the confidence coefficient of the key point groups, and selecting the key point group with the highest confidence coefficient as a reference key point group P i Other key point groups as target key point group P j ;
The elimination process comprises the following steps: judging target key point group P j Whether or not to be deleted, if the key point group P i And target key point group P j If the spatial distance of the target key point group P is not more than the set threshold value j Deleting;
selecting the key point group with highest confidence from the rest target key point groups as a reference key point group P i Other target key point group is P j Repeating the above elimination process untilThe target key point group is empty;
wherein, the key point group P i And target key point group P j The spatial distance of (2) is calculated by:
target key point group P j N-th key point of the key point group P i The spatial distance of the corresponding nth key point isStatistics of target keypoint group P j 8 key points and key point group P i The sum of the spatial distances of the corresponding 8 key points is taken as a key point group P i And target key point group P j Is a spatial distance of (2);
σ 2 for the spatial distance redundancy elimination parameter,representing a set of keypoints P i Coordinates of the nth key point.
6. The method for identifying dislocation faults of bearing saddle of railway wagon according to claim 1, wherein the contour detection model comprises a block unit, a gradient lifting tree classifier and a comprehensive unit, and the contour detection model is used for obtaining the contour of the bearing saddle in an image, and the method comprises the following steps:
the blocking unit blocks the reinforced bearing saddle image by adopting a sliding window to obtain a plurality of image blocks, the image blocks are input into a gradient lifting tree classifier, the gradient lifting tree classifier outputs labels of all the image blocks, the labels of all the image blocks are input into a comprehensive unit, and the comprehensive unit obtains the contour of the bearing saddle in the image according to the labels.
7. The method for identifying dislocation faults of bearing saddles of railway wagons according to claim 1, wherein training a contour detection model comprises:
the method comprises the steps of partitioning an enhanced bearing saddle image by adopting a sliding window, obtaining the characteristics of each image block, calculating the label of each image block by adopting a PCA algorithm, taking the characteristics of the image block as input and the label as output, and forming a training set to train the gradient lifting tree classifier.
8. The method for identifying a misalignment fault of a railroad freight car bearing saddle according to claim 1, wherein the method for obtaining the characteristics of each image block comprises:
processing the image block to enable each pixel to have K-dimensional characteristics, wherein the K-dimensional characteristics comprise color channel characteristics, gradient amplitude channel characteristics, high-pass filtering channel characteristics, maximum fuzzy filtering channel characteristics and Gabor filtering channel characteristics with different dimensions in multiple directions;
and performing dimension reduction on the image block characteristics to obtain dimension-reduced image block characteristics.
9. The method of claim 1, further comprising creating a keypoint detection sample dataset, training a keypoint detection model using the keypoint detection sample dataset, wherein the method of removing red frames from the oblique faulty bearing saddle image when creating the keypoint detection sample dataset comprises:
setting a threshold Th, extracting the region where the red frame is located, and obtaining a white Mask image after large-core morphological expansion;
and repairing a white area in the white Mask image by using an image repairing model lama so as to remove a red frame in the inclined fault bearing saddle image.
10. The method for identifying the dislocation faults of the bearing saddle of the railway wagon according to claim 2, wherein the key point detection sample data set is a data set obtained after sample amplification, and the amplification method comprises the following steps: sharpening, contrast enhancement, rotation, translation, scaling, and mirroring of images.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310615378.1A CN116580245B (en) | 2023-05-29 | 2023-05-29 | Rail wagon bearing saddle dislocation fault identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310615378.1A CN116580245B (en) | 2023-05-29 | 2023-05-29 | Rail wagon bearing saddle dislocation fault identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116580245A true CN116580245A (en) | 2023-08-11 |
CN116580245B CN116580245B (en) | 2023-12-26 |
Family
ID=87543892
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310615378.1A Active CN116580245B (en) | 2023-05-29 | 2023-05-29 | Rail wagon bearing saddle dislocation fault identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116580245B (en) |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105260744A (en) * | 2015-10-08 | 2016-01-20 | 北京航空航天大学 | Automatic on-line diagnosis method for freight train coupler tail cotter position faults and system |
CN109165541A (en) * | 2018-05-30 | 2019-01-08 | 北京飞鸿云际科技有限公司 | Coding method for vehicle component in intelligent recognition rail traffic vehicles image |
WO2019232894A1 (en) * | 2018-06-05 | 2019-12-12 | 中国石油大学(华东) | Complex scene-based human body key point detection system and method |
CN111091557A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Method and system for detecting breakage fault of flange of bearing saddle of railway wagon |
CN111091548A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon adapter dislocation fault image identification method and system based on deep learning |
CN112102280A (en) * | 2020-09-11 | 2020-12-18 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loosening and loss faults of small part bearing key nut of railway wagon |
US20210158104A1 (en) * | 2019-11-26 | 2021-05-27 | Board Of Regents, The University Of Texas System | Using synthetic data sets to train a neural network for three-dimensional seismic fault segmentation |
CN113221839A (en) * | 2021-06-02 | 2021-08-06 | 哈尔滨市科佳通用机电股份有限公司 | Automatic truck image identification method and system |
CN113830136A (en) * | 2021-10-20 | 2021-12-24 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying malposition fault of angle cock handle of railway wagon |
CN114399472A (en) * | 2021-12-28 | 2022-04-26 | 国能铁路装备有限责任公司 | Adapter fault detection method and device, computer equipment and storage medium |
KR102396169B1 (en) * | 2021-09-07 | 2022-05-11 | 주식회사 로드텍 | Apparatus and method for evaluating state of road pavement |
WO2022236647A1 (en) * | 2021-05-11 | 2022-11-17 | Huawei Technologies Co., Ltd. | Methods, devices, and computer readable media for training a keypoint estimation network using cgan-based data augmentation |
-
2023
- 2023-05-29 CN CN202310615378.1A patent/CN116580245B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105260744A (en) * | 2015-10-08 | 2016-01-20 | 北京航空航天大学 | Automatic on-line diagnosis method for freight train coupler tail cotter position faults and system |
CN109165541A (en) * | 2018-05-30 | 2019-01-08 | 北京飞鸿云际科技有限公司 | Coding method for vehicle component in intelligent recognition rail traffic vehicles image |
WO2019232894A1 (en) * | 2018-06-05 | 2019-12-12 | 中国石油大学(华东) | Complex scene-based human body key point detection system and method |
US20210158104A1 (en) * | 2019-11-26 | 2021-05-27 | Board Of Regents, The University Of Texas System | Using synthetic data sets to train a neural network for three-dimensional seismic fault segmentation |
CN111091557A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Method and system for detecting breakage fault of flange of bearing saddle of railway wagon |
CN111091548A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon adapter dislocation fault image identification method and system based on deep learning |
CN112102280A (en) * | 2020-09-11 | 2020-12-18 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loosening and loss faults of small part bearing key nut of railway wagon |
WO2022236647A1 (en) * | 2021-05-11 | 2022-11-17 | Huawei Technologies Co., Ltd. | Methods, devices, and computer readable media for training a keypoint estimation network using cgan-based data augmentation |
CN113221839A (en) * | 2021-06-02 | 2021-08-06 | 哈尔滨市科佳通用机电股份有限公司 | Automatic truck image identification method and system |
KR102396169B1 (en) * | 2021-09-07 | 2022-05-11 | 주식회사 로드텍 | Apparatus and method for evaluating state of road pavement |
CN113830136A (en) * | 2021-10-20 | 2021-12-24 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying malposition fault of angle cock handle of railway wagon |
CN114399472A (en) * | 2021-12-28 | 2022-04-26 | 国能铁路装备有限责任公司 | Adapter fault detection method and device, computer equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
丁春嵘等: "车轮扁疤状态下的铁路货车动力学表征研究", 《机械科学与技术》, vol. 40, no. 08 * |
何步斌;: "厂修货车转K6型承载鞍组装常见故障及处理", 南方职业教育学刊, no. 06 * |
Also Published As
Publication number | Publication date |
---|---|
CN116580245B (en) | 2023-12-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111079747B (en) | Railway wagon bogie side frame fracture fault image identification method | |
CN211731391U (en) | Intelligent online detection device for track surface defects | |
CN104915963B (en) | A kind of detection and localization method for PLCC elements | |
CN111091544B (en) | Method for detecting breakage fault of side integrated framework of railway wagon bogie | |
CN110210350A (en) | A kind of quick parking space detection method based on deep learning | |
CN111899288A (en) | Tunnel leakage water area detection and identification method based on infrared and visible light image fusion | |
CN111311567A (en) | Method for identifying fastener and steel rail diseases of track line image | |
CN107045634A (en) | A kind of text positioning method based on maximum stable extremal region and stroke width | |
CN111091548B (en) | Railway wagon adapter dislocation fault image identification method and system based on deep learning | |
CN116757990A (en) | Railway fastener defect online detection and identification method based on machine vision | |
CN107180416A (en) | Train wheel tread image deformity correction method and system | |
CN111079734B (en) | Method for detecting foreign matters in triangular holes of railway wagon | |
CN111080621B (en) | Method for identifying railway wagon floor damage fault image | |
CN103473570A (en) | Self-adaptive video scene pedestrian detection method | |
CN110222661B (en) | Feature extraction method for moving target identification and tracking | |
CN113830136B (en) | Method for identifying malposition fault of angle cock handle of railway wagon | |
CN112101182A (en) | Railway wagon floor damage fault identification method based on improved SLIC method | |
CN111563896A (en) | Image processing method for catenary anomaly detection | |
CN111738228A (en) | Multi-view vehicle feature matching method for hypermetrological evidence chain verification | |
CN110348307B (en) | Path edge identification method and system for crane metal structure climbing robot | |
CN112712552A (en) | Fault detection method for vehicle tread scratch | |
CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device | |
CN116580245B (en) | Rail wagon bearing saddle dislocation fault identification method | |
CN111833323B (en) | Image quality judgment method for task-divided rail wagon based on sparse representation and SVM (support vector machine) | |
CN112508908B (en) | Method for detecting disconnection fault of sanding pipe joint of motor train unit based on image processing |
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 |