CN116152211A - Identification method for brake shoe abrasion overrun fault - Google Patents
Identification method for brake shoe abrasion overrun fault Download PDFInfo
- Publication number
- CN116152211A CN116152211A CN202310176878.XA CN202310176878A CN116152211A CN 116152211 A CN116152211 A CN 116152211A CN 202310176878 A CN202310176878 A CN 202310176878A CN 116152211 A CN116152211 A CN 116152211A
- Authority
- CN
- China
- Prior art keywords
- brake shoe
- loss function
- image
- overrun
- identifying
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06T7/10—Segmentation; Edge detection
-
- 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
- G06T2207/30204—Marker
-
- 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
- G06T2207/30248—Vehicle exterior or interior
Abstract
A method for identifying brake shoe abrasion overrun faults relates to the technical field of vehicle fault identification. The invention aims to solve the problem that the existing image processing method is difficult to judge the brake shoe abrasion overrun and the identification requirement is met. The invention marks the brake shoe by utilizing the rectangular detection frame on the detected image, and ensures that the center coordinate of the brake shoe is positioned on the diagonal line of the rectangular detection frame; intercepting a rectangular detection frame and rotating the rectangular detection frame anticlockwise; marking the outline of the brake shoe in the rotated image to obtain detection data; inputting the detection data into a trained yoloV5 network to identify faults and obtaining fault identification results; the convolution output layer of the partition network in the yoloV5 network is a deformable convolution output layer, and the loss function of the yoloV5 network is the average value of the BCE loss function, the Struct loss function and the Dice loss function.
Description
Technical Field
The invention belongs to the technical field of vehicle fault identification, and particularly relates to a fault identification method based on image processing.
Background
In the railway safety direction, the traditional method is to find out the fault point of the train through manual observation after the detection equipment takes a picture. This approach allows fault detection while the vehicle is traveling without stopping. However, the use of manual observation has the disadvantages of easy fatigue, high strength, need of training and the like. And the brake shoe abrasion overrun judgment belongs to fine judgment, and the residual width of the brake shoe is difficult to observe by human eyes. Because more and more things can be replaced by machines, and the machines have the characteristics of low cost, unified rule and no fatigue in 24 hours, the image recognition technology is used for replacing the traditional manual detection at the present stage. However, the conventional image processing method has high requirements on the image, and is difficult to meet the accuracy requirements of recognition.
Disclosure of Invention
The invention aims to solve the problem that the existing image processing method is difficult to judge the brake shoe abrasion overrun and the identification requirement is met, and provides a method for identifying the brake shoe abrasion overrun fault.
A method for identifying brake shoe abrasion overrun faults comprises the following steps:
step one: inputting the detected image into a trained yoloV5 network, marking a brake shoe by using a rectangular detection frame by using the yoloV5 network, wherein the center coordinate of the brake shoe is positioned on the diagonal line of the rectangular detection frame;
step two: then intercepting and rotating the rectangular detection frame anticlockwise to obtain a rotated image, wherein the rotation angle theta is as follows:
θ＝arctan(h/w)，
wherein h and w are the height and width of the rectangular detection frame respectively;
step three: marking the outline of the brake shoe in the rotated image to obtain target segmentation data;
step four: inputting the target segmentation data into a trained Unet segmentation network to identify faults, and obtaining a fault identification result;
the convolution output layer of the Unet segmentation network is a deformable convolution output layer, and the loss function of the Unet segmentation network is the average value of the BCE loss function, the Struct loss function and the Dice loss function.
Further, the method for obtaining the measured image comprises the following steps:
and acquiring an image of the detected vehicle, intercepting a bogie area in the image as a sub-image according to the wheelbase information of the detected vehicle, adjusting the gray value of the sub-image to 30, and then increasing the brightness of the sub-image to 90 to obtain the detected image.
Further, in the first step, labeling of the rectangular detection frame is achieved on the detected image by using labelImg.
Further, in the second step, the center coordinates of the brake shoe in the rotated image
The method comprises the following steps:
wherein, (x, y) is the center coordinates of the brake shoe in the rectangular detection frame.
Further, in the third step, the contour of the brake shoe is marked by labelme.
Further, in the fourth step, the LOSS function LOSS expression of the Unet split network is:
LOSS＝(BCELoss+StrctLoss+DiceLoss)/3，
wherein BCELoss is a BCE loss function, strctLoss is a Struct loss function, and DiceLoss is a Dice loss function.
Further, the expression of the BCE loss function BCELoss is as follows:
the expression of the Struct loss function StrctLoss is:
the expression of the Dice loss function DiceLoss is:
above, W n To detect the targetCategory weight, y i Is a real label, y is when the real label is a target i =1, y when the real tag is not the target i ＝0，f(x i ) For the prediction probability of a detection target, i=1, 2,..n, N is the total number of target categories, δ xy Covariance of x and y, and x and y represent true labels and predicted results, delta, respectively x Delta for predicting the average value of the block y For the average value of the label block, c=0.001, tp is the positive number of samples predicted by the model as positive class, FN is the positive number of samples predicted by the model as negative class, and FP is the negative number of samples predicted by the model as positive class.
Further, the method for identifying the brake shoe abrasion overrun fault further comprises the following steps:
step five: judging whether the fault recognition result recognizes the brake shoe, if yes, executing the step six, otherwise outputting a fault alarm signal,
step six: and calculating the horizontal pixel value of the brake shoe according to the output brake shoe image, judging whether the horizontal pixel value is lower than the standard threshold value, if so, outputting a wear alarm signal, otherwise, determining that the brake shoe is normal.
A computer readable storage device storing a computer program which when executed implements a method of identifying a brake shoe wear overrun fault as described above.
The system for identifying the brake shoe wear overrun fault comprises a storage device, a processor and a computer program which is stored in the storage device and can run on the processor, and is characterized in that the processor executes the computer program to realize the method for identifying the brake shoe wear overrun fault.
The beneficial effects of the invention are as follows:
1. the invention replaces manual detection by utilizing an automatic image recognition mode, can solve the fatigue problem of repeated image viewing of manual detection for a long time, unifies standards for the same faults, and improves the detection efficiency and the accuracy.
2. The mixed loss function adopted by the invention can be combined with the target structural characteristics, so that the accuracy of the segmentation edge can be increased.
3. And the DCN is adopted to replace the traditional convolution, so that the segmentation accuracy is improved.
4. And the target is rotated, so that the accuracy of dividing the edge is improved.
Drawings
FIG. 1 is a measured image of a brake shoe marked with a rectangular detection frame;
FIG. 2 is an image of the rotated image marking the contours of the brake shoe;
FIG. 3 is a schematic diagram of the deformable convolution principle of the Unet section;
FIG. 4 is a flow chart of a method for identifying brake shoe wear overrun faults.
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 first embodiment is as follows: referring to fig. 1 to 4, a method for identifying an overrun brake wear failure according to the present embodiment includes the following steps:
step one: inputting a detected image into a trained yolo V5 network, wherein the yolo V5 network marks a brake shoe by utilizing a rectangular detection frame, and the center coordinate of the brake shoe is positioned on the diagonal line of the rectangular detection frame;
step two: intercepting and rotating the rectangular detection frame anticlockwise to obtain a rotated image, wherein the rotation angle theta is as follows:
θ＝arctan(h/w)，
wherein h and w are the height and width of the rectangular detection frame respectively;
step three: marking the outline of the brake shoe in the rotated image to obtain target segmentation data;
step four: inputting the target segmentation data into a trained Unet segmentation network to identify faults and obtain a fault identification result;
the convolution output layer of the Unet segmentation network is a deformable convolution output layer.
In this embodiment, the original convolved output layer is replaced with a Deformable Convolution (DCN); the Unet network is a coding and decoding structure, the original data is subjected to 4 times of downsampling and 4 times of upsampling, the upsampling and downsampling feature map adopts a connection operation to fuse the features of high resolution and low resolution, and the 3*3 convolution layer of the last layer of the network is replaced by a deformable convolution DCN. The deformable convolution can be better adapted to the position of the feature points, and has better effect on non-rectangular targets.
The loss function of the Unet segmentation network is the average value of the BCE loss function, the Struct loss function and the Dice loss function.
The existing loss function is an SSIM (structural similarity) loss function, and is commonly used for image quality assessment by considering brightness, contrast and structural indexes. In quality assessment, the brightness contrast may characterize an image, but for segmentation results, the target gray value is single, and other loss functions are mixed, removing redundancy only preserves the structural comparison partial loss.
In this embodiment, since the identification standard is a pixel level for the problem of brake shoe wear overrun, the target split edge accuracy has a large influence on the identification result. The hybrid loss functions include BCE loss functions (cross entropy loss functions for classification), struct loss functions (structural similarity loss), and Dice loss functions. Wherein the BCE loss function is pixel level loss, the Struct loss function is regional level loss function, and the difference between the prediction target and the real target value structure is judged to be more sensitive to edges, and the Dice loss function is regional level loss function. The combination of the three can remove the situation that the gradient change of the regional level loss function is severe, and can also improve the importance of the target edge. The mixed loss function calculates the loss for each pixel point according to the BCE loss function of the segmentation result, the Strct loss function is the regional block loss, and only the structure comparison part is reserved and the brightness contrast is removed. The Dice loss function is the overall loss. The three loss functions combine to make the target edge clearer.
The second embodiment is as follows: the present embodiment further describes the method for identifying a brake shoe wear overrun fault according to the first embodiment, wherein the method for obtaining the measured image includes:
when the truck passes through a TFDS (truck operation fault dynamic image detection system) detection station, a linear array camera image can be obtained, namely an image of a tested vehicle is obtained, a bogie area is intercepted in the image as a sub-image according to the wheelbase information of the tested vehicle, the gray value of the sub-image is adjusted to 30, and then the brightness of the sub-image is increased to 90, so that the tested image is obtained.
And a third specific embodiment: the embodiment is to further explain the method for identifying the brake shoe wear overrun fault according to the first embodiment, and in the embodiment, labelimg is a rectangular marking tool, which is commonly used for target identification and target detection. In the first step, labeling of a rectangular detection frame is achieved on the detected image by using labelImg, as shown in fig. 1.
The specific embodiment IV is as follows: the method for identifying an overrun brake shoe wear fault according to the first embodiment is further described in this embodiment, wherein in the second step, the center coordinates of the brake shoe in the rotated image
The method comprises the following steps:
wherein, (x, y) is the center coordinates of the brake shoe in the rectangular detection frame.
Fifth embodiment: in this embodiment, compared with the inclined target, the horizontal and vertical target segmentation effect is better, so that the brake shoe abrasion overrun fault identification method according to the first embodiment is used for capturing and rotating according to the marked coordinate point. labelme is a polygonal labeling tool, which can accurately label the outline and is commonly used for segmentation. In step three, the brake shoe profile is marked with labelme, as shown in FIG. 2.
Specific embodiment six: in this embodiment, in the fourth embodiment, the LOSS function LOSS expression of the Unet-split network is:
LOSS＝(BCELoss+StrctLoss+DiceLoss)/3，
wherein BCELoss is a BCE loss function, strctLoss is a Struct loss function, and DiceLoss is a Dice loss function.
Seventh embodiment: the present embodiment is a method for identifying an overrun brake shoe wear failure as described in the first embodiment, in the present embodiment,
the expression of BCE loss function BCELoss is:
the expression of the Struct loss function StrctLoss is:
the expression of the Dice loss function DiceLoss is:
above, W n To detect the class weight of the target, y i Is a real label, y is when the real label is a target i =1, y when the real tag is not the target i ＝0，f(x i ) For the prediction probability of a detection target, i=1, 2,..n, N is the total number of target categories, δ xy As the covariance of x and y,and x and y represent true labels and predicted results, delta, respectively x Delta for predicting the average value of the block y For the average value of the label block, c=0.001, tp is the positive number of samples predicted by the model as positive class, FN is the positive number of samples predicted by the model as negative class, and FP is the negative number of samples predicted by the model as positive class.
Eighth embodiment: the present embodiment further describes a method for identifying an overrun brake shoe wear fault according to the first embodiment, and the method further includes:
step five: judging whether the fault recognition result recognizes the brake shoe, if yes, executing the step six, otherwise outputting a fault alarm signal,
step six: and calculating the horizontal pixel value of the brake shoe according to the output brake shoe image, judging whether the horizontal pixel value is lower than the standard threshold value, if so, outputting a wear alarm signal, otherwise, determining that the brake shoe is normal.
Detailed description nine: a computer-readable storage device according to this embodiment, the storage device storing a computer program that when executed implements the method according to any one of embodiments one to eight.
Detailed description ten: the system for identifying brake shoe wear overrun faults according to the embodiment comprises a storage device, a processor and a computer program stored in the storage device and capable of running on the processor, and is characterized in that the processor executes the computer program to realize the method according to any one of the specific embodiments one to eight.
In summary, the high-speed linear array camera and the acquisition control system are installed at the track side detection station, and when the truck group passes through the detection station, the 2D linear array image and the hardware wheelbase information are obtained. The comprehensive of the data sample is ensured by collecting the driving images under different weather, different time periods and different months at the detection sites in different areas, so that preparation is made for the subsequent model. Because the cameras are consistent, the shooting range and the size change little, the sub-images of the bogie area can be intercepted according to hardware information and priori knowledge. And collecting images of different sites, calculating the gray value of the images, and increasing the brightness of the images to enable the target to be more clearly visible when the integral gray value of the images is smaller than 30. And then intercepting bogie area images, and training the YoloV5 recognition network. According to the invention, the rotation angle is calculated through the coordinate information obtained by the target detection network to rotate the image, and the loss function is optimized to increase the accuracy of the network segmentation edge. The identification program is arranged to the identification server, and when the server receives the signal of the passing car, the identification program is started. When the fault exists in the overtaking, the alarm information is output and uploaded to the platform for manual confirmation and alarm.
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. The method for identifying the brake shoe abrasion overrun fault is characterized by comprising the following steps of:
step one: inputting a detected image into a trained yolo V5 network, wherein the yolo V5 network marks a brake shoe by utilizing a rectangular detection frame, and the center coordinate of the brake shoe is positioned on the diagonal line of the rectangular detection frame;
step two: intercepting and rotating the rectangular detection frame anticlockwise to obtain a rotated image, wherein the rotation angle theta is as follows:
θ＝arctan(hw)，
wherein h and w are the height and width of the rectangular detection frame respectively;
step three: marking the outline of the brake shoe in the rotated image to obtain target segmentation data;
step four: inputting the target segmentation data into a trained Unet segmentation network to identify faults and obtain a fault identification result;
the convolutional output layer of the said Unet split network is a deformable convolutional output layer,
the loss function of the Unet segmentation network is the mean value of the BCE loss function, the Struct loss function and the Dice loss function.
2. The method for identifying brake shoe wear overrun faults according to claim 1, wherein the method for obtaining the measured image is as follows:
and acquiring an image of the detected vehicle, intercepting a bogie area in the image as a sub-image according to the wheelbase information of the detected vehicle, adjusting the gray value of the sub-image to 30, and then increasing the brightness of the sub-image to 90 to obtain the detected image.
3. A method for identifying an overrun brake shoe wear failure in accordance with claim 1, wherein,
in the first step, the label of the rectangular detection frame is realized on the detected image by using labelImg.
4. A method for identifying an overrun brake shoe wear failure in accordance with claim 1, wherein,
in the second step, the center coordinates of the brake shoe in the rotated image
The method comprises the following steps:
wherein, (x, y) is the center coordinates of the brake shoe in the rectangular detection frame.
5. A method for identifying an overrun brake shoe wear failure in accordance with claim 1, wherein,
and thirdly, marking the outline of the brake shoe by using labelme.
6. A method for identifying an overrun brake shoe wear failure in accordance with claim 1, wherein,
in the fourth step, the LOSS function LOSS expression of the said Unet split network is:
LOSS＝(BCELoss+StrctLoss+DiceLoss)3，
wherein BCELoss is a BCE loss function, strctLoss is a Struct loss function, and DiceLoss is a Dice loss function.
7. The method for identifying a brake shoe wear out fault as claimed in claim 6, wherein,
the expression of the BCE loss function BCELoss is:
the expression of the Struct loss function StrctLoss is:
the expression of the Dice loss function DiceLoss is:
above, W n To detect the class weight of the target, y i Is a real label, y is when the real label is a target i =1, y when the real tag is not the target i ＝0，f(x i ) For the prediction probability of a detection target, i=1, 2,..n, N is the total number of target categories, δ xy Covariance of x and y, and x and y represent true labels and predicted results, delta, respectively x Delta for predicting the average value of the block y For the average value of the label block, c=0.001, tp is the positive number of samples predicted by the model as positive class, FN is the positive number of samples predicted by the model as negative class, and FP is the negative number of samples predicted by the model as positive class.
8. The method for identifying brake shoe wear out fault as defined in claim 1, further comprising:
step five: judging whether the fault recognition result recognizes the brake shoe, if yes, executing the step six, otherwise outputting a fault alarm signal,
step six: and calculating the horizontal pixel value of the brake shoe according to the output brake shoe image, judging whether the horizontal pixel value is lower than the standard threshold value, if so, outputting a wear alarm signal, otherwise, determining that the brake shoe is normal.
9. A computer-readable storage device storing a computer program, characterized in that the computer program when executed implements the method according to any one of claims 1 to 8.
10. A brake shoe wear out fault identification system comprising a storage device, a processor and a computer program stored in the storage device and executable on the processor, wherein execution of the computer program by the processor implements the method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310176878.XA CN116152211A (en) | 2023-02-28 | 2023-02-28 | Identification method for brake shoe abrasion overrun fault |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310176878.XA CN116152211A (en) | 2023-02-28 | 2023-02-28 | Identification method for brake shoe abrasion overrun fault |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116152211A true CN116152211A (en) | 2023-05-23 |
Family
ID=86356711
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310176878.XA Pending CN116152211A (en) | 2023-02-28 | 2023-02-28 | Identification method for brake shoe abrasion overrun fault |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116152211A (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5327782A (en) * | 1991-09-09 | 1994-07-12 | Kawasaki Steel Corporation | Automatic brake shoe measuring apparatus for rolling stock |
| CN106989672A (en) * | 2017-04-17 | 2017-07-28 | 天津大学 | A kind of workpiece measuring based on machine vision |
| CN109191476A (en) * | 2018-09-10 | 2019-01-11 | 重庆邮电大学 | The automatic segmentation of Biomedical Image based on U-net network structure |
| US20190063533A1 (en) * | 2017-08-29 | 2019-02-28 | Kawasaki Jukogyo Kabushiki Kaisha | Brake shoe abrasion detection system of railway vehicle |
| CN111080668A (en) * | 2019-12-13 | 2020-04-28 | 武汉华目信息技术有限责任公司 | Brake pad wear fault detection method and system |
| US20200327660A1 (en) * | 2019-04-10 | 2020-10-15 | International Business Machines Corporation | Automated fracture detection using machine learning models |
| CN113592822A (en) * | 2021-08-02 | 2021-11-02 | 郑州大学 | Insulator defect positioning method for power inspection image |
| CN114399672A (en) * | 2022-01-14 | 2022-04-26 | 东南大学 | Railway wagon brake shoe fault detection method based on deep learning |
| CN114782471A (en) * | 2022-04-12 | 2022-07-22 | 首都医科大学附属北京天坛医院 | Method for segmenting ultrasonic two-dimensional image of thyroid nodule |
| CN114943721A (en) * | 2022-06-07 | 2022-08-26 | 哈尔滨理工大学 | Neck ultrasonic image segmentation method based on improved U-Net network |
| CN114998852A (en) * | 2021-08-05 | 2022-09-02 | 浙江杉工智能科技有限公司 | Intelligent detection method for road pavement diseases based on deep learning |
| CN114998373A (en) * | 2022-06-15 | 2022-09-02 | 南京信息工程大学 | Improved U-Net cloud picture segmentation method based on multi-scale loss function |
| CN115331086A (en) * | 2022-08-17 | 2022-11-11 | 哈尔滨市科佳通用机电股份有限公司 | Brake shoe breaking and rivet losing fault detection method |
| CN115546664A (en) * | 2022-09-30 | 2022-12-30 | 湖北省电力勘测设计院有限公司 | Cascaded network-based insulator self-explosion detection method and system |
| CN115661144A (en) * | 2022-12-15 | 2023-01-31 | 湖南工商大学 | Self-adaptive medical image segmentation method based on deformable U-Net |
-
2023
- 2023-02-28 CN CN202310176878.XA patent/CN116152211A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5327782A (en) * | 1991-09-09 | 1994-07-12 | Kawasaki Steel Corporation | Automatic brake shoe measuring apparatus for rolling stock |
| CN106989672A (en) * | 2017-04-17 | 2017-07-28 | 天津大学 | A kind of workpiece measuring based on machine vision |
| US20190063533A1 (en) * | 2017-08-29 | 2019-02-28 | Kawasaki Jukogyo Kabushiki Kaisha | Brake shoe abrasion detection system of railway vehicle |
| CN109191476A (en) * | 2018-09-10 | 2019-01-11 | 重庆邮电大学 | The automatic segmentation of Biomedical Image based on U-net network structure |
| US20200327660A1 (en) * | 2019-04-10 | 2020-10-15 | International Business Machines Corporation | Automated fracture detection using machine learning models |
| CN111080668A (en) * | 2019-12-13 | 2020-04-28 | 武汉华目信息技术有限责任公司 | Brake pad wear fault detection method and system |
| CN113592822A (en) * | 2021-08-02 | 2021-11-02 | 郑州大学 | Insulator defect positioning method for power inspection image |
| CN114998852A (en) * | 2021-08-05 | 2022-09-02 | 浙江杉工智能科技有限公司 | Intelligent detection method for road pavement diseases based on deep learning |
| CN114399672A (en) * | 2022-01-14 | 2022-04-26 | 东南大学 | Railway wagon brake shoe fault detection method based on deep learning |
| CN114782471A (en) * | 2022-04-12 | 2022-07-22 | 首都医科大学附属北京天坛医院 | Method for segmenting ultrasonic two-dimensional image of thyroid nodule |
| CN114943721A (en) * | 2022-06-07 | 2022-08-26 | 哈尔滨理工大学 | Neck ultrasonic image segmentation method based on improved U-Net network |
| CN114998373A (en) * | 2022-06-15 | 2022-09-02 | 南京信息工程大学 | Improved U-Net cloud picture segmentation method based on multi-scale loss function |
| CN115331086A (en) * | 2022-08-17 | 2022-11-11 | 哈尔滨市科佳通用机电股份有限公司 | Brake shoe breaking and rivet losing fault detection method |
| CN115546664A (en) * | 2022-09-30 | 2022-12-30 | 湖北省电力勘测设计院有限公司 | Cascaded network-based insulator self-explosion detection method and system |
| CN115661144A (en) * | 2022-12-15 | 2023-01-31 | 湖南工商大学 | Self-adaptive medical image segmentation method based on deformable U-Net |
Non-Patent Citations (2)
| Title |
|---|
| RONG ZOU;ZHEN-YING XU;JIN-YANG LI;FU-QIANG ZHOU;: "铁路货车闸瓦钎故障的实时监控（英文）", JOURNAL OF ZHEJIANG UNIVERSITY-SCIENCE C(COMPUTERS & ELECTRONICS), vol. 1995, no. 03, pages 228 - 229 * |
| 杨雪荣;高向东;成思源;: "基于计算机视觉的列车闸瓦检测方法", 内燃机车, no. 06 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110211101A (en) | A kind of rail surface defect rapid detection system and method | |
| CN109489724B (en) | Tunnel train safe operation environment comprehensive detection device and detection method | |
| CN111080600A (en) | Fault identification method for split pin on spring supporting plate of railway wagon | |
| CN110599469A (en) | Method and system for detecting defects of key parts of motor train unit and electronic equipment | |
| CN113516629A (en) | Intelligent detection system for TFDS passing operation | |
| CN106503640A (en) | A kind of detection method for taking bus zone | |
| CN102610104B (en) | Onboard front vehicle detection method | |
| CN111080650A (en) | Method for detecting looseness and loss faults of small part bearing blocking key nut of railway wagon | |
| CN111754460A (en) | Method, system and storage medium for automatically detecting gap of point switch | |
| CN113788051A (en) | Train on-station running state monitoring and analyzing system | |
| CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device | |
| Wang et al. | License plate location algorithm based on edge detection and morphology | |
| CN111080599A (en) | Fault identification method for hook lifting rod of railway wagon | |
| Doycheva et al. | Computer vision and deep learning for real-time pavement distress detection | |
| CN112037182B (en) | Locomotive running part fault detection method and device based on time sequence image and storage medium | |
| CN112329858B (en) | Image recognition method for breakage fault of anti-loosening iron wire of railway motor car | |
| CN112233095A (en) | Method for detecting multiple fault forms of railway wagon locking plate device | |
| CN112085723A (en) | Automatic detection method for spring jumping fault of truck bolster | |
| CN116152211A (en) | Identification method for brake shoe abrasion overrun fault | |
| CN115170923B (en) | Fault identification method for loss of railway wagon supporting plate nut | |
| KR102585146B1 (en) | Pothole detection device and method based on deep learning | |
| CN115424128A (en) | Fault image detection method and system for lower link of freight car bogie | |
| CN109978879B (en) | Box corner in-groove state detection method based on railway wagon loading video monitoring | |
| CN112949644B (en) | Fault image identification method for lower pull rod of truck | |
| Yang et al. | An automatic defect detection system based on deep learning for fasteners in the power system |
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 |