CN113205133B - Tunnel water stain intelligent identification method based on multitask learning - Google Patents
Tunnel water stain intelligent identification method based on multitask learning Download PDFInfo
- Publication number
- CN113205133B CN113205133B CN202110482829.XA CN202110482829A CN113205133B CN 113205133 B CN113205133 B CN 113205133B CN 202110482829 A CN202110482829 A CN 202110482829A CN 113205133 B CN113205133 B CN 113205133B
- Authority
- CN
- China
- Prior art keywords
- image
- wet
- detection
- classification
- slices
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 14
- 238000001514 detection method Methods 0.000 claims abstract description 81
- 230000007547 defect Effects 0.000 claims abstract description 8
- 230000005540 biological transmission Effects 0.000 claims abstract description 5
- 238000004806 packaging method and process Methods 0.000 claims abstract description 4
- 238000007635 classification algorithm Methods 0.000 claims description 42
- 238000004422 calculation algorithm Methods 0.000 claims description 35
- 238000013135 deep learning Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 240000004282 Grewia occidentalis Species 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/24765—Rule-based classification
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a tunnel water stain intelligent identification method based on multi-task learning, which comprises the following steps: image acquisition and positioning; collecting wet-stain image data of a subway tunnel, and simultaneously obtaining positioning information data of the wet-stain image through a positioning module; image classification, identification and detection; the method comprises the steps of receiving collected image data, transmitting the data to a multi-task learning identification module, detecting and classifying wet spots of a subway tunnel in real time to obtain detection results and classification results respectively, carrying out result combination on the detection results and the classification results, and outputting identification results; packaging the defect picture, the final identification result and the found positioning information into an alarm file according to the identification result; the vehicle-mounted transmission program transmits the packaged alarm file to the data terminal through a network; the data terminal analyzes the alarm file and then is used for relevant staff to check and maintain in time. The invention can reduce the labor burden and the time cost and improve the working efficiency of railway workers.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a tunnel water stain intelligent recognition method based on multi-task learning.
Background
Tunnel detection is an important means of construction quality management and is a precondition for ensuring train operation safety. The tunnel wet-spot can influence the normal running of the traffic flow in the tunnel and even jeopardize the safe running of the train. For existing tunnel wet stains, defect detection is generally carried out in a manual inspection or offline data analysis mode, but the traditional modes are time-consuming and labor-consuming, potential safety hazards cannot be eliminated timely, great trouble is brought to railway workers, but a single vehicle-mounted recognition detection algorithm leaks part of the wet stains, and detection recall is not high, so that the system provides a recognition mode based on multi-task learning to realize real-time, intelligent and efficient recognition of the tunnel wet stains.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a tunnel water stain intelligent identification method based on multi-task learning, which comprises the following steps:
image acquisition and positioning; collecting wet-stain image data of a subway tunnel, and simultaneously obtaining positioning information data of the wet-stain image through a positioning module;
image classification, identification and detection; the method comprises the steps of receiving collected image data, transmitting the data to a multi-task learning identification module, detecting and classifying wet spots of a subway tunnel in real time to obtain detection results and classification results respectively, carrying out result combination on the detection results and the classification results, and outputting identification results;
packaging the defect picture, the final identification result and the found positioning information into an alarm file according to the identification result; the vehicle-mounted transmission program transmits the packaged alarm file to the data terminal through a network; the data terminal analyzes the alarm file and then is used for relevant staff to check and maintain in time.
The intelligent recognition method for tunnel water stains based on multi-task learning according to claim 1, wherein the steps of collecting the wet stain image data of the subway tunnel and simultaneously obtaining the positioning information data of the wet stain image through a positioning module comprise the following steps:
in a good satellite searching area, using GPS to perform initial positioning, and in a pilot blind area, using a base station to perform initial positioning; the inertial navigation device is utilized to perform continuous positioning, and meanwhile, when GPS signals or base station signals exist, position correction is performed; and synchronizing the positioning information and the picture clock information, and associating the nearest neighbor positioning information for each frame of image.
Further, the receiving the collected image data, transmitting the data to a multi-task learning identification module, and detecting the wet spot of the subway tunnel in real time, comprising the following steps:
slicing the acquired images before detection, namely slicing the acquired images at the positions of four corners of each image and cutting out slices with the same size as the slices at the positions of the four corners by taking the central point of each image as an origin, wherein each image corresponds to five slice images; preprocessing the image after slicing by using a deep learning method, resetting the size of the image and normalizing the image;
and (3) inputting the preprocessed images into a detection network by adopting a yolov3 detection algorithm, wherein the detection algorithm detects slices positioned at four corners of each image: scaling the input image to the specified size of the detection network, extracting the characteristics and obtaining a characteristic diagram;
the feature map is divided into NxN cells, and the center coordinates of the object are predicted from which cell the center coordinates of the object fall.
Further, the collecting the collected image data, transmitting the data to a multi-task learning identification module, and classifying the wet stains of the subway tunnel, comprising the following steps:
and inputting the normalized image into a classification network by adopting a classification algorithm, wherein the classification algorithm identifies and classifies four slices at the same position and slices at the middle position of the same image, and the classification network adopts an acceptance_v3 model to extract and classify the characteristics of the image.
Further, the step of combining the detection result and the classification result to output the identification result comprises the following steps:
a) For the slices positioned at the four corners of the original image, if the detection algorithm and the classification algorithm both recognize wet stains or the detection algorithm detects wet stains but the classification algorithm does not recognize wet stains, outputting a detection result of the detection algorithm, wherein the detection result comprises position information and confidence of the wet stains;
b) If the detection algorithm does not detect wet stains on the slices at the same position and the classification algorithm identifies the wet stains, outputting an identification result of the classification algorithm;
c) If the detection algorithm and the classification algorithm do not detect wet stains on the four corner slices, and the classification algorithm identifies wet stains on the middle slice, outputting an identification result of the classification algorithm;
if the detection algorithm or the classification algorithm identifies wet stains on two or more slices, merging the identified results on the slices corresponding to the same frame image, and outputting a final identification result.
The beneficial effects of the invention are as follows: and detecting and identifying the tunnel wet stain by using a detection program and a classification program, and merging the detection result and the identification result in real time and returning to the alarm terminal. The whole system is deployed on NVIDIA Jetson AGX Xavier, so that wet stains in a tunnel can be identified rapidly and efficiently in real time, potential hazards can be eliminated by railway workers in time, safe running of a train is ensured, meanwhile, labor burden and time cost are reduced, and working efficiency of the railway workers is improved.
Drawings
FIG. 1 is a flow chart of a method for intelligently identifying water stains in a tunnel based on multi-task learning;
FIG. 2 is a schematic view of an image slice;
FIG. 3 is a schematic flow chart of the yolov3 algorithm;
fig. 4 is a schematic diagram of a classification process of the classification algorithm.
Detailed Description
The technical solution of the present invention will be described in further detail with reference to the accompanying drawings, but the scope of the present invention is not limited to the following description.
As shown in fig. 1, a tunnel water stain intelligent identification method based on multi-task learning comprises the following steps:
image acquisition and positioning; collecting wet-stain image data of a subway tunnel, and simultaneously obtaining positioning information data of the wet-stain image through a positioning module;
image classification, identification and detection; the method comprises the steps of receiving collected image data, transmitting the data to a multi-task learning identification module, detecting and classifying wet spots of a subway tunnel in real time to obtain detection results and classification results respectively, carrying out result combination on the detection results and the classification results, and outputting identification results;
packaging the defect picture, the final identification result and the found positioning information into an alarm file according to the identification result; the vehicle-mounted transmission program transmits the packaged alarm file to the data terminal through a network; the data terminal analyzes the alarm file and then is used for relevant staff to check and maintain in time.
The intelligent recognition method for tunnel water stains based on multi-task learning according to claim 1, wherein the steps of collecting the wet stain image data of the subway tunnel and simultaneously obtaining the positioning information data of the wet stain image through a positioning module comprise the following steps:
in a good satellite searching area, using GPS to perform initial positioning, and in a pilot blind area, using a base station to perform initial positioning; the inertial navigation device is utilized to perform continuous positioning, and meanwhile, when GPS signals or base station signals exist, position correction is performed; and synchronizing the positioning information and the picture clock information, and associating the nearest neighbor positioning information for each frame of image.
The method comprises the following steps of receiving collected image data, transmitting the data to a multi-task learning identification module, and detecting the wet spots of the subway tunnel in real time, wherein the method comprises the following steps:
slicing the acquired images before detection, namely slicing the acquired images at the positions of four corners of each image and cutting out slices with the same size as the slices at the positions of the four corners by taking the central point of each image as an origin, wherein each image corresponds to five slice images; preprocessing the image after slicing by using a deep learning method, resetting the size of the image and normalizing the image;
and (3) inputting the preprocessed images into a detection network by adopting a yolov3 detection algorithm, wherein the detection algorithm detects slices positioned at four corners of each image: scaling the input image to the specified size of the detection network, extracting the characteristics and obtaining a characteristic diagram;
the feature map is divided into NxN cells, and the center coordinates of the object are predicted from which cell the center coordinates of the object fall.
The collected image data is transmitted to a multi-task learning identification module to classify the wet stains of the subway tunnel, and the method comprises the following steps:
and inputting the normalized image into a classification network by adopting a classification algorithm, wherein the classification algorithm identifies and classifies four slices at the same position and slices at the middle position of the same image, and the classification network adopts an acceptance_v3 model to extract and classify the characteristics of the image.
The detection result and the classification result are combined as a result, and the identification result is output, and the method comprises the following steps:
a) For the slices positioned at the four corners of the original image, if the detection algorithm and the classification algorithm both recognize wet stains or the detection algorithm detects wet stains but the classification algorithm does not recognize wet stains, outputting a detection result of the detection algorithm, wherein the detection result comprises position information and confidence of the wet stains;
b) If the detection algorithm does not detect wet stains on the slices at the same position and the classification algorithm identifies the wet stains, outputting an identification result of the classification algorithm;
c) If the detection algorithm and the classification algorithm do not detect wet stains on the four corner slices, and the classification algorithm identifies wet stains on the middle slice, outputting an identification result of the classification algorithm;
if the detection algorithm or the classification algorithm identifies wet stains on two or more slices, merging the identified results on the slices corresponding to the same frame image, and outputting a final identification result.
Specifically, the method adopts a design scheme of on-line intelligent recognition of tunnel wet stains based on multi-task learning, utilizes a detection program and a classification program to detect and recognize subway tunnel wet stains simultaneously, and returns a recognition detection result to an alarm terminal to generate a detection report.
And acquiring image data required by detecting the wet stain of the subway tunnel and obtaining positioning information. The data acquisition module mainly comprises a camera and light supplementing, and images wet spots of the tunnel in real time. The positioning information is obtained through a base station, a GPS and an inertial navigation combined positioning module, and the specific positioning steps are as follows:
1) In a good satellite searching area, using GPS to perform initial positioning, and in a pilot blind area, using a base station to perform initial positioning;
2) The inertial navigation device is utilized to perform continuous positioning, and meanwhile, when a GPS signal or a base station signal exists, position correction is performed to prevent position drift;
3) And synchronizing positioning and picture clock information, and associating nearest neighbor position information for each frame of picture.
Image data acquired by a camera is received, the data are transmitted to a deployed defect detection and classification program, and the wet spots of the subway tunnel are detected and classified in real time.
1) The detection process is as follows:
i. before the detection, the detected data is subjected to slicing processing, for example, the length and width of original pictures are 5120, and 3000x3000 slices are cut at the four corners of each picture, namely, each picture corresponds to four slice diagrams, as shown in fig. 1.
Preprocessing the image data after slicing by a deep learning method, resetting the image size (resetting the input image size by using an opencv library self-contained reset function) in order to enable the received data to meet the input requirement of a deep learning detection method, and normalizing the image data for better detection.
To enable rapid detection, the yolov3 detection algorithm was used. The backbone network employed by yolov3 is dark net53. The detection flow is approximately as follows:
after inputting data into the detection network:
scaling the input image to the specified size of the detection network, extracting the characteristics and obtaining a characteristic diagram with a certain size;
the feature map is divided into NxN cells, and if the center coordinates of the object fall in which cell, the cell predicts the object and the positional information.
i. The classification algorithm and the detection algorithm simultaneously identify slices at the same frame and the same position, but the classification algorithm identifies one more slice at the middle position: a 3000x3000 slice was cut with the center point of each picture as the origin, as shown in fig. 1.
Preprocessing the image data after slicing by using a deep learning method, resetting the image size (resetting the input image size by using an opencv library self-contained resetting function) in order to enable the received data to meet the input requirement of a classification method, and normalizing the image data for better classification.
inputting the normalized image into a classification network, and carrying out feature extraction and classification on the image by the classification network by adopting an acceptance_v3 model.
3) Detection task and classification task result combination
The detection algorithm detects the slices positioned at four corners of each frame of picture, and at the same time, the classification algorithm identifies and classifies the four slices positioned at the same position of the same frame of picture and the slices positioned in the middle position:
a) For the slices positioned at the four corners of the original image, if the detection algorithm and the classification algorithm both recognize wet stains, the detection algorithm detects the wet stains but the classification algorithm does not recognize the wet stains, outputting the detection result (comprising the position information and the confidence of the wet stains) of the detection algorithm;
b) If the detection algorithm does not detect wet stains on the slices at the same position and the classification algorithm identifies the wet stains, the identification result of the classification algorithm (the position of the wet stains, namely the position information and the confidence of the slices corresponding to the original image) is output;
c) If the detection algorithm and the classification algorithm do not detect wet stains on the four-corner slices, and the classification algorithm identifies wet stains on the middle slice, outputting the identification result of the classification algorithm, namely the position information and the confidence of the slice corresponding to the original image;
if the detection algorithm or the classification algorithm identifies wet stains on two or more slices, merging the identified results on the slices corresponding to the same frame of picture, and outputting a final identification result.
The terminal alarm module packages the defect picture, the final identification result and the found positioning information into an alarm file; and the vehicle-mounted transmission program transmits the packaged alarm file to the data terminal through the 4G network.
The foregoing is merely a preferred embodiment of the invention, and it is to be understood that the invention is not limited to the form disclosed herein but is not to be construed as excluding other embodiments, but is capable of numerous other combinations, modifications and environments and is capable of modifications within the scope of the inventive concept, either as taught or as a matter of routine skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.
Claims (4)
1. The intelligent tunnel water stain identification method based on multi-task learning is characterized by comprising the following steps of:
image acquisition and positioning; collecting wet-stain image data of a subway tunnel, and simultaneously obtaining positioning information data of the wet-stain image through a positioning module;
image classification, identification and detection; the method comprises the steps of receiving collected image data, transmitting the data to a multi-task learning identification module, detecting and classifying wet spots of a subway tunnel in real time to obtain detection results and classification results respectively, carrying out result combination on the detection results and the classification results, and outputting identification results;
packaging the defect picture, the final identification result and the found positioning information into an alarm file according to the identification result; the vehicle-mounted transmission program transmits the packaged alarm file to the data terminal through a network; the data terminal analyzes the alarm file and then is used for related staff to check and maintain in time;
the detection result and the classification result are combined as a result, and the identification result is output, and the method comprises the following steps:
a) For the slices positioned at the four corners of the original image, if the detection algorithm and the classification algorithm both recognize wet stains or the detection algorithm detects wet stains but the classification algorithm does not recognize wet stains, outputting a detection result of the detection algorithm, wherein the detection result comprises position information and confidence of the wet stains;
b) If the detection algorithm does not detect wet stains on the slices at the same position and the classification algorithm identifies the wet stains, outputting an identification result of the classification algorithm;
c) If the detection algorithm and the classification algorithm do not detect wet stains on the four corner slices, and the classification algorithm identifies wet stains on the middle slice, outputting an identification result of the classification algorithm;
if the detection algorithm or the classification algorithm identifies wet stains on two or more slices, merging the identified results on the slices corresponding to the same frame image, and outputting a final identification result.
2. The intelligent recognition method for tunnel water stains based on multi-task learning according to claim 1, wherein the steps of collecting the wet stain image data of the subway tunnel and simultaneously obtaining the positioning information data of the wet stain image through a positioning module comprise the following steps:
in a good satellite searching area, using GPS to perform initial positioning, and in a pilot blind area, using a base station to perform initial positioning; the inertial navigation device is utilized to perform continuous positioning, and meanwhile, when GPS signals or base station signals exist, position correction is performed; and synchronizing the positioning information and the picture clock information, and associating the nearest neighbor positioning information for each frame of image.
3. The intelligent recognition method of tunnel water stain based on multi-task learning according to claim 1, wherein the steps of receiving the collected image data, transmitting the data to a multi-task learning recognition module, and detecting the wet stain of the subway tunnel in real time comprise the following steps:
slicing the acquired images before detection, namely slicing the acquired images at the positions of four corners of each image and cutting out slices with the same size as the slices at the positions of the four corners by taking the central point of each image as an origin, wherein each image corresponds to five slice images; preprocessing the image after slicing by using a deep learning method, resetting the size of the image and normalizing the image;
and (3) inputting the preprocessed images into a detection network by adopting a yolov3 detection algorithm, wherein the detection algorithm detects slices positioned at four corners of each image: scaling the input image to the specified size of the detection network, extracting the characteristics and obtaining a characteristic diagram;
the feature map is divided into NxN cells, and the center coordinates of the object are predicted from which cell the center coordinates of the object fall.
4. The intelligent recognition method of tunnel water stains based on multi-task learning according to claim 3, wherein the collected image data is transmitted to a multi-task learning recognition module to classify the subway tunnel water stains, and the method comprises the following steps:
and inputting the normalized image into a classification network by adopting a classification algorithm, wherein the classification algorithm identifies and classifies four slices at the same position and slices at the middle position of the same image, and the classification network adopts an acceptance_v3 model to extract and classify the characteristics of the image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110482829.XA CN113205133B (en) | 2021-04-30 | 2021-04-30 | Tunnel water stain intelligent identification method based on multitask learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110482829.XA CN113205133B (en) | 2021-04-30 | 2021-04-30 | Tunnel water stain intelligent identification method based on multitask learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113205133A CN113205133A (en) | 2021-08-03 |
| CN113205133B true CN113205133B (en) | 2024-01-26 |
Family
ID=77030139
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110482829.XA Active CN113205133B (en) | 2021-04-30 | 2021-04-30 | Tunnel water stain intelligent identification method based on multitask learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113205133B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117010831B (en) * | 2023-08-07 | 2024-03-08 | 深圳远大科技工程有限公司 | Building intelligent management system and method based on big data |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160096460A (en) * | 2015-02-05 | 2016-08-16 | 삼성전자주식회사 | Recognition system based on deep learning including a plurality of classfier and control method thereof |
| EP3273387A1 (en) * | 2016-07-19 | 2018-01-24 | Siemens Healthcare GmbH | Medical image segmentation with a multi-task neural network system |
| CN109325454A (en) * | 2018-09-28 | 2019-02-12 | 合肥工业大学 | A kind of static gesture real-time identification method based on YOLOv3 |
| KR20190032908A (en) * | 2017-09-20 | 2019-03-28 | 주식회사 에이치엔에스휴먼시스템 | Method for managing steel quality and system |
| CN109754399A (en) * | 2019-01-17 | 2019-05-14 | 成都国铁电气设备有限公司 | A kind of railway power supply detection repetition of alarms analysis method based on image segmentation |
| CN109977957A (en) * | 2019-03-04 | 2019-07-05 | 苏宁易购集团股份有限公司 | A kind of invoice recognition methods and system based on deep learning |
| CN110119726A (en) * | 2019-05-20 | 2019-08-13 | 四川九洲视讯科技有限责任公司 | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model |
| CN111369515A (en) * | 2020-02-29 | 2020-07-03 | 上海交通大学 | Tunnel water stain detection system and method based on computer vision |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| CN111932548A (en) * | 2020-04-21 | 2020-11-13 | 南京航空航天大学 | Subway tunnel fine crack segmentation method based on deep learning |
| CN112070754A (en) * | 2020-09-11 | 2020-12-11 | 武汉百家云科技有限公司 | Tunnel segment water leakage detection method and device, electronic equipment and medium |
| CN112113978A (en) * | 2020-09-22 | 2020-12-22 | 成都国铁电气设备有限公司 | Vehicle-mounted tunnel defect online detection system and method based on deep learning |
| CN112115902A (en) * | 2020-09-25 | 2020-12-22 | 广州市派客朴食信息科技有限责任公司 | Dish identification method based on single-stage target detection algorithm |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11170504B2 (en) * | 2019-05-02 | 2021-11-09 | Keyamed Na, Inc. | Method and system for intracerebral hemorrhage detection and segmentation based on a multi-task fully convolutional network |
-
2021
- 2021-04-30 CN CN202110482829.XA patent/CN113205133B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160096460A (en) * | 2015-02-05 | 2016-08-16 | 삼성전자주식회사 | Recognition system based on deep learning including a plurality of classfier and control method thereof |
| EP3273387A1 (en) * | 2016-07-19 | 2018-01-24 | Siemens Healthcare GmbH | Medical image segmentation with a multi-task neural network system |
| KR20190032908A (en) * | 2017-09-20 | 2019-03-28 | 주식회사 에이치엔에스휴먼시스템 | Method for managing steel quality and system |
| CN109325454A (en) * | 2018-09-28 | 2019-02-12 | 合肥工业大学 | A kind of static gesture real-time identification method based on YOLOv3 |
| CN109754399A (en) * | 2019-01-17 | 2019-05-14 | 成都国铁电气设备有限公司 | A kind of railway power supply detection repetition of alarms analysis method based on image segmentation |
| CN109977957A (en) * | 2019-03-04 | 2019-07-05 | 苏宁易购集团股份有限公司 | A kind of invoice recognition methods and system based on deep learning |
| CN110119726A (en) * | 2019-05-20 | 2019-08-13 | 四川九洲视讯科技有限责任公司 | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model |
| CN111369515A (en) * | 2020-02-29 | 2020-07-03 | 上海交通大学 | Tunnel water stain detection system and method based on computer vision |
| CN111932548A (en) * | 2020-04-21 | 2020-11-13 | 南京航空航天大学 | Subway tunnel fine crack segmentation method based on deep learning |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| CN112070754A (en) * | 2020-09-11 | 2020-12-11 | 武汉百家云科技有限公司 | Tunnel segment water leakage detection method and device, electronic equipment and medium |
| CN112113978A (en) * | 2020-09-22 | 2020-12-22 | 成都国铁电气设备有限公司 | Vehicle-mounted tunnel defect online detection system and method based on deep learning |
| CN112115902A (en) * | 2020-09-25 | 2020-12-22 | 广州市派客朴食信息科技有限责任公司 | Dish identification method based on single-stage target detection algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度可分离卷积的地铁隧道巡检视频分析;孙明华;杨媛;李渊博;;计算机工程与科学(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113205133A (en) | 2021-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP4109331A1 (en) | Obstacle detection method and apparatus, computer device, and storage medium | |
| CN112113978A (en) | Vehicle-mounted tunnel defect online detection system and method based on deep learning | |
| KR102346676B1 (en) | Method for creating damage figure using the deep learning-based damage image classification of facility | |
| CN109584222A (en) | A kind of failure modes of the photovoltaic module image based on unmanned plane and discrimination method | |
| CN101834989B (en) | Real-time data acquisition and storage system of helicopter in electric inspection process | |
| CN110509951A (en) | A kind of rail deformation detection system and method | |
| CN109993163A (en) | A kind of non-rating plate identifying system and its recognition methods based on artificial intelligence | |
| WO2019198076A1 (en) | Real-time raw data- and sensor fusion | |
| CN111833302A (en) | Car number and case number identification association system | |
| CN110956104A (en) | Method, device and system for detecting overflow of garbage can | |
| CN113205133B (en) | Tunnel water stain intelligent identification method based on multitask learning | |
| CN112528979B (en) | Transformer substation inspection robot obstacle distinguishing method and system | |
| CN112184773A (en) | Helmet wearing detection method and system based on deep learning | |
| CN111161292A (en) | Ore size measurement method and application system | |
| CN110969610A (en) | Power equipment infrared chart identification method and system based on deep learning | |
| CN114882440A (en) | Human head detection method and system | |
| CN113935364A (en) | Unmanned aerial vehicle flame detection method and system based on YOLOv4 network | |
| CN111339927B (en) | Intelligent recognition system for working state of personnel in electric power business hall | |
| CN117726958B (en) | Intelligent detection and hidden danger identification method for inspection image target of unmanned aerial vehicle of distribution line | |
| CN111126112A (en) | Candidate region determination method and device | |
| CN113092495A (en) | Intelligent inspection system and method for subway tunnel defects with cooperation of train and ground | |
| CN115620000A (en) | Intelligent identification method for accumulated water of slag-free ballast bed | |
| CN212302511U (en) | Car number and case number identification association system monitoring facilities | |
| CN116597501A (en) | Video analysis algorithm and edge device | |
| CN113569943A (en) | Deep neural network-based slag piece bulk early warning method, system and device |
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 |