CN112183342A - Comprehensive convertor station defect identification method with template - Google Patents
Comprehensive convertor station defect identification method with template Download PDFInfo
- Publication number
- CN112183342A CN112183342A CN202011041010.1A CN202011041010A CN112183342A CN 112183342 A CN112183342 A CN 112183342A CN 202011041010 A CN202011041010 A CN 202011041010A CN 112183342 A CN112183342 A CN 112183342A
- Authority
- CN
- China
- Prior art keywords
- picture
- detected
- detection
- frames
- hidden danger
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Abstract
The invention discloses a comprehensive converter station defect identification method with a template, which comprises the steps of 1, acquiring a reference picture and a picture to be detected in a converter station sample library, wherein the sample library is a set of reference pictures, the reference picture refers to a normal picture when a converter station has no hidden danger, the picture to be detected is a picture of the same converter station position shot in real time, 2, obtaining a difference map of the reference picture and the picture to be detected in the sample library by synthesizing the simultaneous comparison results of two or more detection algorithms, and 3, prompting the possible position of the hidden danger according to the difference map. The invention can effectively improve the defect identification accuracy of the existing converter station, reduce the possibility of false alarm and missed alarm and reduce the burden of workers.
Description
Technical Field
The invention relates to the technical field of converter station equipment, in particular to a comprehensive converter station defect identification method with a template.
Background
The converter station plays an extremely important role as an important part in a power system, but because the equipment is numerous and complicated and the structure of most of the equipment is complicated, the comprehensive and accurate defect detection of the equipment is extremely difficult. At present, defect inspection of the converter station is mainly carried out in a manual inspection mode, the method is time-consuming and labor-consuming, and false inspection and missed inspection often occur due to the influence of the number of personnel and the quality of the personnel. Even if a few converter stations pass through intelligent detection equipment, an online patrol system is established, the participation of manpower is reduced, and the actual use requirement cannot be met. The reason is that the defect detection method of the general online patrol system is greatly reduced in accuracy under the characteristics of complex environmental conditions and various equipment types of the converter station, and cannot realize comprehensive and accurate defect identification.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a comprehensive convertor station defect identification method with a template, so that the accuracy of convertor station defect identification and detection can be improved, the possibility of misinformation and missing report can be reduced, and the burden of workers can be reduced, thereby improving the overall efficiency of a convertor station and reducing the possibility of large-scale accidents.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention discloses a method for identifying defects of a comprehensive convertor station with a template, which is characterized by comprising the following steps of:
step 2, detecting the picture to be detected by using an SSIM algorithm based on the reference picture to obtain an SSIM detection result;
detecting the to-be-detected picture by using a machine vision detection algorithm of deep learning to obtain a vision algorithm detection result;
step 3, judging whether the similarity value of the reference picture and the picture to be detected in the SSIM detection result is larger than a set similarity limit value or not, if so, indicating that the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected; otherwise, executing step 4;
step 4, judging whether a similar judgment frame exists in the SSIM detection result, if so, executing step 5, otherwise, indicating that the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected;
step 5, acquiring the number of similar judgment frames, and if the number of the similar judgment frames is 1, executing step 6; if the number of the similar judgment frames is two or more, executing the step 7;
step 6, judging the detection result of the visual algorithm, and outputting a conclusion according to the judgment result;
step 7, judging whether a detection frame exists in the detection result of the visual algorithm, and if so, executing step 8; otherwise, directly taking the display position of the similarity judging frame on the picture to be detected as a hidden danger position for prompting;
step 8, judging whether the number of the detection frames is 1, if so, executing step 9; otherwise, when the number of the detection frames is two or more, executing the step 10;
step 9, judging the priority type of the detection frame, and outputting a corresponding conclusion according to a judgment result;
step 10, judging whether the types of the detection frames contain priority types, if so, directly taking the display positions of all the detection frames on the picture to be detected as hidden danger positions to prompt, otherwise, executing step 11;
and 11, calculating the contact ratio of all the similar judgment frames and all the detection frames respectively, and if the obtained contact ratio average value reaches a preset threshold value D, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of all the similar judgment frames on the picture to be detected as hidden danger positions.
The method for identifying the defects of the comprehensive converter station is also characterized in that the step 6 is carried out according to the following steps:
6.1, judging whether a detection frame exists in the detection result of the visual algorithm, and if so, executing the step 6.2; otherwise, directly taking the display position of the similarity judging frame on the picture to be detected as a hidden danger position for prompting;
step 6.2, acquiring the number of the detection frames, and if the number of the detection frames is 1, executing step 6.3; if the number of the detection frames is two or more, executing the step 6.4;
step 6.3, calculating the contact ratio of the similar judging frame and the detection frame, and when the obtained result reaches a preset threshold value A, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or prompting by taking the display position of the similar judging frame on the picture to be detected as a hidden danger position;
6.4, judging whether the detection frame types contain priority types, if so, prompting the display positions of all the detection frames on the picture to be detected as hidden danger positions, and if not, executing the step 6.5;
and 6.5, calculating the coincidence degrees of the similar judgment frames and all the detection frames, and when the obtained average value of the coincidence degrees reaches a preset threshold value B, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of the similar judgment frames on the picture to be detected as hidden danger positions.
The step 9 is carried out according to the following steps:
9.1, judging whether the type of the detection frame contains a priority type, if so, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, otherwise, executing a step 9.2;
and 9.2, respectively calculating the coincidence degrees of all the similar judgment frames and the detection frame, and when the obtained coincidence degree average value reaches a preset threshold value C, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or otherwise, prompting by taking the display positions of all the judgment frames on the picture to be detected as hidden danger positions.
The priority class refers to a key detection class set in a sample recognition training process by a machine vision detection algorithm based on deep learning.
The magnitude relationship of each threshold is A > B > C > D.
Compared with the prior art, the invention has the beneficial effects that:
the method adopts two different algorithms of an SSIM algorithm and a deep learning computer vision algorithm to simultaneously identify and detect the picture to be detected, integrates the advantages of the two algorithms, and mutually reflects the detection result, so that the accuracy of the final output result is greatly improved, the defect detection and identification accuracy of the converter station is effectively improved, and the possibility of large-scale accident risk is reduced. The improvement of the accuracy of the detection result can increase the working proportion of the online patrol system, effectively reduce the burden of workers, reduce the labor participation in defect patrol, improve the working efficiency and greatly improve the overall benefit of the converter station.
Drawings
FIG. 1 is a flow chart of a specific implementation of the method for identifying defects of the comprehensive converter station with the template;
Detailed Description
In this embodiment, a method for identifying defects of a comprehensive converter station with a template is performed according to the following steps:
step 2, detecting the picture to be detected by using an SSIM algorithm based on the reference picture to obtain an SSIM detection result;
detecting the to-be-detected picture by utilizing a deep learning computer vision algorithm to obtain a vision algorithm detection result;
as shown in fig. 1, SSIMbox is an SSIM algorithm, the detector box refers to a computer vision algorithm for deep learning, the to-be-detected picture and the reference picture are processed by the SSIM algorithm to obtain a comparison similarity value and a difference distinguishing frame between the reference picture and the to-be-detected picture, and the to-be-detected picture passes through the detector box and is output as soon as a hidden danger covered in the training set is detected.
The method adopts two different detection algorithms to process the picture to be detected because a single SSIM algorithm and a deep learning algorithm have larger defects.
Firstly, the SSIM algorithm needs to align a reference picture with a picture to be detected, if a problem occurs in the alignment process, the subsequent comparison result is greatly influenced, secondly, when a complex scene is processed, the SSIM algorithm result is easily interfered, thirdly, the change of the shadow position of an object caused by illumination change can be detected, and fourthly, the semantic information of the SSIM algorithm needs to be processed independently. However, the SSIM algorithm does not require a large amount of preliminary work, does not consider whether a sample to be detected appears before, and only needs to compare a reference picture with a picture to be detected.
The deep learning algorithm does not need to be aligned with the reference picture, and only a large number of sample pictures are needed for training; sample information can be effectively identified in a complex scene; hardly affected by the light and shadow transform; there is no need to process the semantic information separately. However, the deep learning algorithm has low accuracy in processing the sample information which does not appear, and is easy to have false detection and missing detection.
Therefore, the detection accuracy can be greatly improved by combining the advantages of the two algorithms to process the obtained result.
Step 3, judging whether the similarity value between the reference picture and the picture to be detected in the SSIM detection result is greater than a set similarity limit value, wherein the set similarity limit value is 0.995 as shown in figure 1, and if so, the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected; otherwise, executing step 4;
step 4, judging whether a similarity judgment frame exists in the SSIM detection result, if so, executing step 5, otherwise, indicating that the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected;
step 5, acquiring the number of similar decision frames, and if the number of similar decision frames is 1, that is, if the len (ssimbox) is 1 as shown in fig. 1, determining that the number is Y, executing step 6; if the number of the similar decision frames is two or more, that is, the judgment result is N when len (ssimbox) is 1, then step 7 is executed;
step 6, judging the detection result of the visual algorithm, and outputting a conclusion according to the judgment result;
step 6, shown by the dashed box in fig. 1, is performed as follows:
6.1, judging whether a detection frame exists in a visual algorithm detection result, and if so, executing the step 6.2; if the hidden danger does not exist, the detector box does not detect the hidden danger, and the display position of the similar judgment frame on the picture to be detected is directly used as the hidden danger position for prompting;
step 6.2, acquiring the number of detection frames, and if the number of detection frames is 1, that is, if the len (netbox) is 1 as shown in the dotted line frame in fig. 1, determining that the number is Y, executing step 6.3; if the number of the detection frames is two or more, that is, if len (netbox) 1 is judged to be N, then step 6.4 is executed;
step 6.3, calculating the contact ratio of the similar judging frame and the detection frame, and when the obtained result reaches a preset threshold value A, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or prompting by taking the display position of the similar judging frame on the picture to be detected as a hidden danger position;
as shown in fig. 1, the threshold condition is that the IOU is greater than 0.5, when the coincidence degree result meets the condition, the result shown by the default detection frame is correct, and the display position of the detection frame on the picture to be detected is used as a hidden danger position for prompting; and when the conditions are not met, the display area of the detection frame is the same as the reference picture, the result shown by the detection frame is wrong, and the display position of the similar judgment frame on the picture to be detected is used as a hidden danger position for prompting.
6.4, judging whether the detection frame types contain priority types, if so, prompting the display positions of all the detection frames on the picture to be detected as hidden danger positions, and if not, executing the step 6.5;
the priority class refers to a key detection class set in a sample recognition training process by a machine vision detection algorithm based on deep learning. The key detection type is the type of potential safety hazard which has great influence on the converter station, and the principle of 'no discharge after error killing' is taken as a priority judgment condition. Therefore, the display positions of all the detection frames on the picture to be detected are directly used as hidden danger positions to prompt without calculating the contact ratio.
And 6.5, calculating the coincidence degrees of the similar judgment frames and all the detection frames, and when the obtained coincidence degree average value reaches a preset threshold value B, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of the similar judgment frames on the picture to be detected as hidden danger positions.
The threshold condition is IOU >0.4 as shown in the dashed box of FIG. 1, because there is more than one detection box, the deviation is relatively large when the coincidence degree calculation is performed with the judgment box, and the set threshold condition is lower than the threshold A.
Step 7, judging whether a detection frame exists in the detection result of the visual algorithm, and if so, executing step 8; otherwise, directly taking the display position of the similarity judging frame on the picture to be detected as a hidden danger position for prompting;
step 8, judging whether the number of the detection frames is 1, if so, executing step 9; otherwise, when the number of the detection frames is two or more, executing the step 10;
step 9, judging the priority type of the detection frame, and outputting a corresponding conclusion according to a judgment result;
step 9, shown by the dotted box in fig. 1, is performed as follows:
9.1, judging whether the type of the detection frame contains a priority type, if so, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, otherwise, executing a step 9.2;
and 9.2, respectively calculating the coincidence degrees of all the similar judgment frames and the detection frame, and when the obtained coincidence degree average value reaches a preset threshold value C, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or otherwise, prompting by taking the display positions of all the judgment frames on the picture to be detected as hidden danger positions.
The threshold condition is IOU >0.3 as shown in the dotted line box of FIG. 1, because there is more than one decision box at this time, and the SSIM algorithm has a slightly lower accuracy of the detection result than the deep learning computer vision algorithm under the complex scene condition, the threshold condition is set to be lower than the threshold B.
Step 10, judging whether the types of the detection frames contain priority types, if so, directly taking the display positions of all the detection frames on the picture to be detected as hidden danger positions to prompt, otherwise, executing step 11;
and 11, calculating the contact ratio of all the similar judgment frames and all the detection frames respectively, and if the obtained contact ratio average value reaches a preset threshold value D, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of all the similar judgment frames on the picture to be detected as hidden danger positions.
As shown in fig. 1, the threshold condition is IOU >0.25, and at this time, the number of detection frames and number of determination frames are two or more, which greatly affects the result of the calculation of the degree of coincidence, so the set threshold condition is the lowest.
Claims (5)
1. A comprehensive convertor station defect identification method with a template is characterized by comprising the following steps:
step 1, obtaining a reference picture and a to-be-detected picture in a converter station sample library, wherein the reference picture is a normal picture when no hidden trouble exists in the converter station, and the to-be-detected picture is a picture of the same converter station position shot in real time;
step 2, detecting the picture to be detected by using an SSIM algorithm based on the reference picture to obtain an SSIM detection result;
detecting the to-be-detected picture by using a machine vision detection algorithm of deep learning to obtain a vision algorithm detection result;
step 3, judging whether the similarity value of the reference picture and the picture to be detected in the SSIM detection result is larger than a set similarity limit value or not, if so, indicating that the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected; otherwise, executing step 4;
step 4, judging whether a similar judgment frame exists in the SSIM detection result, if so, executing step 5, otherwise, indicating that the picture to be detected is the same as a normal picture, namely, no hidden danger is detected in the picture to be detected;
step 5, acquiring the number of similar judgment frames, and if the number of the similar judgment frames is 1, executing step 6; if the number of the similar judgment frames is two or more, executing the step 7;
step 6, judging the detection result of the visual algorithm, and outputting a conclusion according to the judgment result;
step 7, judging whether a detection frame exists in the detection result of the visual algorithm, and if so, executing step 8; otherwise, directly taking the display position of the similarity judging frame on the picture to be detected as a hidden danger position for prompting;
step 8, judging whether the number of the detection frames is 1, if so, executing step 9; otherwise, when the number of the detection frames is two or more, executing the step 10;
step 9, judging the priority type of the detection frame, and outputting a corresponding conclusion according to a judgment result;
step 10, judging whether the types of the detection frames contain priority types, if so, directly taking the display positions of all the detection frames on the picture to be detected as hidden danger positions to prompt, otherwise, executing step 11;
and 11, calculating the contact ratio of all the similar judgment frames and all the detection frames respectively, and if the obtained contact ratio average value reaches a preset threshold value D, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of all the similar judgment frames on the picture to be detected as hidden danger positions.
2. The integrated converter station fault identification method according to claim 1, characterized in that said step 6 is performed as follows:
6.1, judging whether a detection frame exists in the detection result of the visual algorithm, and if so, executing the step 6.2; otherwise, directly taking the display position of the similarity judging frame on the picture to be detected as a hidden danger position for prompting;
step 6.2, acquiring the number of the detection frames, and if the number of the detection frames is 1, executing step 6.3; if the number of the detection frames is two or more, executing the step 6.4;
step 6.3, calculating the contact ratio of the similar judging frame and the detection frame, and when the obtained result reaches a preset threshold value A, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or prompting by taking the display position of the similar judging frame on the picture to be detected as a hidden danger position;
6.4, judging whether the detection frame types contain priority types, if so, prompting the display positions of all the detection frames on the picture to be detected as hidden danger positions, and if not, executing the step 6.5;
and 6.5, calculating the coincidence degrees of the similar judgment frames and all the detection frames, and when the obtained average value of the coincidence degrees reaches a preset threshold value B, prompting by taking the display positions of all the detection frames on the picture to be detected as hidden danger positions, or otherwise, prompting by taking the display positions of the similar judgment frames on the picture to be detected as hidden danger positions.
3. The integrated converter station fault identification method according to claim 1, characterized in that said step 9 is performed as follows:
9.1, judging whether the type of the detection frame contains a priority type, if so, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, otherwise, executing a step 9.2;
and 9.2, respectively calculating the coincidence degrees of all the similar judgment frames and the detection frame, and when the obtained coincidence degree average value reaches a preset threshold value C, prompting by taking the display position of the detection frame on the picture to be detected as a hidden danger position, or otherwise, prompting by taking the display positions of all the judgment frames on the picture to be detected as hidden danger positions.
4. The integrated converter station defect identification method of claim 1, wherein the priority class is a key detection class set by a deep learning-based machine vision detection algorithm in a sample identification training process.
5. The method for identifying defects in a comprehensive converter station according to claims 1 to 3, wherein the magnitude relationship of each threshold is A > B > C > D.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011041010.1A CN112183342B (en) | 2020-09-28 | 2020-09-28 | Comprehensive convertor station defect identification method with template |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011041010.1A CN112183342B (en) | 2020-09-28 | 2020-09-28 | Comprehensive convertor station defect identification method with template |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112183342A true CN112183342A (en) | 2021-01-05 |
| CN112183342B CN112183342B (en) | 2022-07-12 |
Family
ID=73943789
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011041010.1A Active CN112183342B (en) | 2020-09-28 | 2020-09-28 | Comprehensive convertor station defect identification method with template |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112183342B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050286753A1 (en) * | 2004-06-25 | 2005-12-29 | Triant Technologies Inc. | Automated inspection systems and methods |
| JP2017097718A (en) * | 2015-11-26 | 2017-06-01 | 株式会社リコー | Identification processing device, identification system, identification method, and program |
| CN106952257A (en) * | 2017-03-21 | 2017-07-14 | 南京大学 | A kind of curved surface label open defect detection method based on template matches and Similarity Measure |
| CN108982508A (en) * | 2018-05-23 | 2018-12-11 | 江苏农林职业技术学院 | A kind of plastic-sealed body IC chip defect inspection method based on feature templates matching and deep learning |
| CN109447154A (en) * | 2018-10-29 | 2019-03-08 | 网易（杭州）网络有限公司 | Picture similarity detection method, device, medium and electronic equipment |
| CN109670515A (en) * | 2018-12-13 | 2019-04-23 | 南京工业大学 | A kind of detection method and system changed for building in unmanned plane image |
| CN110378869A (en) * | 2019-06-05 | 2019-10-25 | 北京交通大学 | A kind of rail fastening method for detecting abnormality of sample automatic marking |
| CN110570536A (en) * | 2019-08-26 | 2019-12-13 | 北京许继电气有限公司 | intelligent line patrol system for extra-high voltage transmission line |
| CN110610485A (en) * | 2019-08-26 | 2019-12-24 | 北京许继电气有限公司 | Ultra-high voltage transmission line channel hidden danger early warning method based on SSIM algorithm |
| CN111024708A (en) * | 2019-09-06 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Method, device, system and equipment for processing product defect detection data |
| CN111340791A (en) * | 2020-03-02 | 2020-06-26 | 浙江浙能技术研究院有限公司 | Photovoltaic module unsupervised defect detection method based on GAN improved algorithm |
| CN111563896A (en) * | 2020-07-20 | 2020-08-21 | 成都中轨轨道设备有限公司 | Image processing method for catenary anomaly detection |
| CN111598857A (en) * | 2020-05-11 | 2020-08-28 | 北京阿丘机器人科技有限公司 | Method and device for detecting surface defects of product, terminal equipment and medium |
-
2020
- 2020-09-28 CN CN202011041010.1A patent/CN112183342B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050286753A1 (en) * | 2004-06-25 | 2005-12-29 | Triant Technologies Inc. | Automated inspection systems and methods |
| JP2017097718A (en) * | 2015-11-26 | 2017-06-01 | 株式会社リコー | Identification processing device, identification system, identification method, and program |
| CN106952257A (en) * | 2017-03-21 | 2017-07-14 | 南京大学 | A kind of curved surface label open defect detection method based on template matches and Similarity Measure |
| CN108982508A (en) * | 2018-05-23 | 2018-12-11 | 江苏农林职业技术学院 | A kind of plastic-sealed body IC chip defect inspection method based on feature templates matching and deep learning |
| CN109447154A (en) * | 2018-10-29 | 2019-03-08 | 网易（杭州）网络有限公司 | Picture similarity detection method, device, medium and electronic equipment |
| CN109670515A (en) * | 2018-12-13 | 2019-04-23 | 南京工业大学 | A kind of detection method and system changed for building in unmanned plane image |
| CN110378869A (en) * | 2019-06-05 | 2019-10-25 | 北京交通大学 | A kind of rail fastening method for detecting abnormality of sample automatic marking |
| CN110570536A (en) * | 2019-08-26 | 2019-12-13 | 北京许继电气有限公司 | intelligent line patrol system for extra-high voltage transmission line |
| CN110610485A (en) * | 2019-08-26 | 2019-12-24 | 北京许继电气有限公司 | Ultra-high voltage transmission line channel hidden danger early warning method based on SSIM algorithm |
| CN111024708A (en) * | 2019-09-06 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Method, device, system and equipment for processing product defect detection data |
| CN111340791A (en) * | 2020-03-02 | 2020-06-26 | 浙江浙能技术研究院有限公司 | Photovoltaic module unsupervised defect detection method based on GAN improved algorithm |
| CN111598857A (en) * | 2020-05-11 | 2020-08-28 | 北京阿丘机器人科技有限公司 | Method and device for detecting surface defects of product, terminal equipment and medium |
| CN111563896A (en) * | 2020-07-20 | 2020-08-21 | 成都中轨轨道设备有限公司 | Image processing method for catenary anomaly detection |
Non-Patent Citations (5)
| Title |
|---|
| BEIXIN XIA等: ""SSIM-NET: Real-Time PCB Defect Detection Based on SSIM and MobileNet-V3"", 《2019 2ND WORLD CONFERENCE ON MECHANICAL ENGINEERING AND INTELLIGENT MANUFACTURING (WCMEIM)》, 20 February 2020 (2020-02-20), pages 756 - 759 * |
| JING WANG等: ""Visual Defect Detection for Substation Equipment based on Joint Inspection Data of Camera and Robot"", 《2020 IEEE 5TH INFORMATION TECHNOLOGY AND MECHATRONICS ENGINEERING CONFERENCE (ITOEC)》, 16 July 2020 (2020-07-16), pages 491 - 495 * |
| 乔羽等: ""基于加权SSIM算法的PCB元件缺陷检测"", 《制造业自动化》, 30 April 2019 (2019-04-30), pages 10 - 14 * |
| 刘萍: ""基于深度学习的视频异常事件检测"", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 January 2020 (2020-01-15), pages 138 - 1240 * |
| 高吉: ""基于改进结构相似度的标签缺陷检测系统研究与实现"", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 October 2018 (2018-10-15), pages 138 - 542 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112183342B (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663360B (en) | Method for automatic identifying steel slab coding and steel slab tracking system | |
| CN110108712A (en) | Multifunctional visual sense defect detecting system | |
| CN111460969A (en) | Intelligent industrial information monitoring system based on cloud computing | |
| WO2022222467A1 (en) | Open circular ring workpiece appearance defect detection method and system, and computer storage medium | |
| CN109086643B (en) | Color box label detection method and system based on machine vision | |
| CN112183342B (en) | Comprehensive convertor station defect identification method with template | |
| CN111626104B (en) | Cable hidden trouble point detection method and device based on unmanned aerial vehicle infrared thermal image | |
| CN111274872B (en) | Video monitoring dynamic irregular multi-supervision area discrimination method based on template matching | |
| CN116310294A (en) | Subway train bolt loosening detection method and device | |
| CN114622311A (en) | Yarn breakage detection method and device and spinning machine | |
| CN115620192A (en) | Method and device for detecting wearing of safety rope in aerial work | |
| KR20190119801A (en) | Vehicle Headlight Alignment Calibration and Classification, Inspection of Vehicle Headlight Defects | |
| CN114742823A (en) | Intelligent detection method for scratches on surface of object | |
| CN113504244A (en) | New energy automobile battery flaw detection method and device | |
| CN112288747A (en) | Intelligent detection method and device for steel billets | |
| CN112329783B (en) | Image processing-based coupler yoke break identification method | |
| CN111223086A (en) | Building crack identification and identification effect optimization method based on deep learning | |
| CN112329775B (en) | Character recognition method for digital multimeter | |
| CN116311073A (en) | Intelligent security system and operation method | |
| CN112465784B (en) | Metro clamp appearance abnormality detection method | |
| CN113962956B (en) | Foreign matter detection method for coal conveying belt conveyor | |
| CN112634352B (en) | Transformer substation fisheye type opening and closing state identification method and system | |
| CN117035669A (en) | Enterprise safety production management method and system based on artificial intelligence | |
| CN116612088A (en) | Defect detection method and system for disassembled parts of waste service end | |
| KR20230024160A (en) | System and method for semantic segmentation learning data error detection and correction |
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 |