CN110443791B - Workpiece detection method and device based on deep learning network - Google Patents
Workpiece detection method and device based on deep learning network Download PDFInfo
- Publication number
- CN110443791B CN110443791B CN201910712712.9A CN201910712712A CN110443791B CN 110443791 B CN110443791 B CN 110443791B CN 201910712712 A CN201910712712 A CN 201910712712A CN 110443791 B CN110443791 B CN 110443791B
- Authority
- CN
- China
- Prior art keywords
- workpiece
- image
- detected
- network
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- 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
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a workpiece detection method and a workpiece detection device based on a deep learning network, which specifically comprise the following steps: acquiring a sample workpiece image, labeling the sample workpiece image, and acquiring position information of the sample workpiece image; collecting a workpiece image to be detected, sending the workpiece image to be detected into an image cache region, and judging whether a workpiece exists in the image cache region or not; if the workpiece exists on the workpiece image to be detected, performing defect detection on the workpiece image to be detected; and if the workpiece does not exist on the image of the workpiece to be detected, stopping detection. The workpiece classification and defect detection can be simultaneously realized, and the method is suitable for flexible detection of workpieces of various types, small batches and frequent product updating.
Description
Technical Field
The invention belongs to the technical field of workpiece detection methods, and relates to a workpiece detection method based on a deep learning network; also relates to a detection device of the workpiece detection method.
Background
The manufacturing industry is the basic industry of national economy, and with the development of science and technology, intelligent manufacturing has become a breakthrough and a necessary way for the transformation from a large manufacturing country to a strong manufacturing country. Industrial robots have become an important mark for measuring the state manufacturing and technology levels as irreplaceable important equipment and means in intelligent manufacturing. The robot industry, which mainly uses industrial robots, is an important path selection for solving the problems of cost rise and environmental restriction in the manufacturing industry. In the intelligent manufacturing industry, the detection of parts by adopting the industrial robot vision technology is favorable for finding the process problems in the production process, improving the production process problems in time, improving the product quality and being favorable for part classification and sorting.
As is known, intelligent manufacturing is required to meet flexible production of various small-batch features, and quality detection of the flexible production is required to be suitable for classification, positioning and feature extraction of parts of different types and detection of surface defects of various features; at present, workpiece detection is mainly completed by means of artificial vision, although the artificial time detection adaptability is strong, the detection of characteristics of various workpieces can be performed, the detection speed is low, and the requirements of online rapid detection are difficult to meet depending on artificial experience; the existing robot has high visual detection speed, but is difficult to adapt to the detection of various workpieces and different characteristics.
Disclosure of Invention
The invention aims to provide a workpiece detection method based on a deep learning network, and solves the problem that the robot visual detection in the prior art cannot adapt to the detection of various workpieces.
The invention adopts the technical scheme that a workpiece detection method based on a deep learning network specifically comprises the following steps:
step 1, collecting a sample workpiece image, labeling the sample workpiece image, and acquiring position information of the sample workpiece image;
step 2, collecting a workpiece image to be detected, sending the workpiece image to be detected into an image cache region, and judging whether a workpiece exists in the image cache region or not;
step 3, if the workpiece exists on the workpiece image to be detected, performing defect detection on the workpiece image to be detected; and if the workpiece does not exist on the image of the workpiece to be detected, stopping detection.
The invention is also characterized in that:
the step 1 specifically comprises the following steps: and collecting a sample workpiece image, and marking the sample workpiece image by an image marking tool to obtain the workpiece category and the pixel point coordinates of the sample workpiece image.
The step 2 specifically comprises the following steps: sending the workpiece image to be detected into an image cache region, drawing a suspicious region suggestion frame on the workpiece image to be detected through a region suggestion network, sending the workpiece image to be detected into a feature pyramid for image segmentation to obtain a suspicious region of the workpiece image to be detected, comparing the suspicious region of the workpiece image to be detected with a sample workpiece image, and judging whether a workpiece exists in the image cache region.
The step 3 specifically comprises the following steps:
step 3.1, inputting the workpiece image to be detected into a residual error network to obtain a characteristic image;
step 3.2, inputting the characteristic image into the area suggestion network to obtain class parameters, class logic parameters and boundary frame parameters, and inhibiting the class parameters, the class logic parameters and the boundary frame parameters through a non-maximum value to obtain an interest area;
and 3.3, inputting the position information of the interest region and the sample workpiece image into a target detection layer of a region suggestion network for target detection to obtain an initial workpiece category, a boundary frame and a segmentation mask, aligning the interest region after combining the workpiece category, the boundary frame, the segmentation mask and the feature image to obtain a feature pyramid network, and obtaining the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected through a positioning network and a segmentation network of the feature pyramid network.
And 3.4, inputting the accurate workpiece category, the workpiece boundary box and the workpiece segmentation mask into an ROI classifier to obtain an initial defect category and a rough defect boundary box, and then obtaining a final mask result through a segmentation mask network.
The invention further aims to provide a workpiece detection device based on the deep learning network.
The invention adopts another technical scheme that a workpiece detection device based on a deep learning network comprises:
the positioning module is used for acquiring a sample workpiece image, labeling the sample workpiece image and obtaining the position information of the sample workpiece image;
the identification module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region or not;
the detection module is used for detecting the defects of the workpiece image to be detected if the workpiece exists on the workpiece image to be detected; and if the workpiece does not exist on the image of the workpiece to be detected, stopping detection.
The positioning module includes:
the first acquisition module is used for acquiring a sample workpiece image;
and the marking module is used for marking the sample workpiece image and acquiring the position information of the sample workpiece image.
The identification module comprises:
the second acquisition module is used for acquiring an image of the workpiece to be detected;
and the judging module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region.
The detection module includes:
the first acquisition module is used for inputting the workpiece image to be detected into a residual error network to acquire a characteristic image;
the second acquisition module is used for inputting the characteristic image into the area suggestion network to obtain class parameters, class logic parameters and boundary frame parameters, and obtaining the interest area after the class parameters, the class logic parameters and the boundary frame parameters are inhibited by a non-maximum value;
the third acquisition module is used for inputting the position information of the interest area and the sample workpiece image into a target detection layer of a region suggestion network for target detection to obtain an initial workpiece category, a boundary frame and a segmentation mask, aligning the interest area after combining the workpiece category, the boundary frame, the segmentation mask and the feature image to obtain a feature pyramid network, and acquiring the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected through a positioning network and a segmentation network of the feature pyramid network;
and the fourth acquisition module is used for inputting the accurate workpiece category, the workpiece boundary box and the workpiece segmentation mask into the ROI classifier to obtain an initial defect category and a rough defect boundary box, and then acquiring a final mask result through a segmentation mask network.
The invention has the beneficial effects that: the workpiece detection method can realize the classification and defect detection of the workpieces at the same time, and is suitable for flexible detection of workpieces of various types, small batches and frequent product updating.
Drawings
FIG. 1 is a flow chart of a workpiece detection method based on a deep learning network according to the present invention;
FIG. 2 is a selected image annotation tool box diagram of the workpiece detection method based on the deep learning network;
FIG. 3 is a schematic diagram of the coordinates of a workpiece obtained by a positioning module of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 4 is a flow chart of a detection module of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 5 is an initial workpiece category, a bounding box, and a segmentation mask obtained by the workpiece detection method based on the deep learning network according to the present invention;
FIG. 6 is a schematic structural diagram of a regional recommendation network of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 7 is an anchor point generated by the area-suggested-network scan of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 8 is a schematic structural diagram of a feature pyramid network of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 9 is a schematic structural diagram of a ROI classifier of the workpiece detection method based on the deep learning network according to the present invention;
FIG. 10 is a diagram of a segmented mask network mask process of a deep learning network-based workpiece inspection method according to the present invention;
fig. 11 is a schematic structural diagram of a workpiece detection apparatus based on a deep learning network according to the present invention.
In the figure, 1 is a positioning module, 101 is a first acquisition module, 102 is a labeling module, 2 is an identification module, 201 is a second acquisition module, 202 is a judgment module, 3 is a detection module, 301 is a first acquisition module, 302 is a second acquisition module, 303 is a third acquisition module, 304 is a fourth acquisition module.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention discloses a workpiece detection method based on a deep learning network, which specifically comprises the following steps as shown in figure 1:
step 1, collecting a sample workpiece image, and labeling the sample workpiece image to obtain position information of the sample workpiece image;
the sample workpiece image is marked by an image marking tool (labelImg), the image marking tool can adopt point selection and frame selection, and the frame selection is selected because the workpiece detected by the invention has an irregular shape, as shown in fig. 2. Then, the pixel point coordinates of the workpiece are obtained through labelImg, as shown in table 1, that is, the position information of the workpiece, including the workpiece type, the workpiece height, the workpiece width of the sample workpiece image, and the maximum y-axis value y of the workpiece coordinates max Y-axis minimum value y min X-axis maximum x max X-axis minimum value x min As shown in fig. 3, to thereby effect positioning of the workpiece.
TABLE 1 coordinate information (Unit: pixel) corresponding to the position of the workpiece
Step 2, collecting a workpiece image to be detected, sending the workpiece image to be detected into an image cache region, and judging whether a workpiece exists in the image cache region or not;
sending a workpiece image to be detected into an image cache region, drawing a suspicious region suggestion frame on the workpiece image to be detected through a region suggestion network, sending the workpiece image to be detected into a feature pyramid for image segmentation, wherein the feature pyramid allows the feature of each level to be combined with the high-level feature and the low-level feature, and the effect of predicting the suspicious region of the image is achieved by fusing the features of different layers, so that the suspicious region of the workpiece image to be detected is obtained; and comparing the suspicious region of the workpiece image to be detected with the sample workpiece image, and judging whether the workpiece exists in the image cache region.
Step 3, if the workpiece exists on the to-be-detected workpiece image, performing defect detection on the to-be-detected workpiece image, as shown in fig. 4; and if the workpiece does not exist on the image of the workpiece to be detected, stopping detection.
Step 3.1, inputting the workpiece image to be detected into a residual error network, and generating a p2, p3, p4, p5 and p6 characteristic image through 5-layer convolution of the residual error network;
step 3.2, inputting the p2, p3, p4, p5 and p6 characteristic images into an area suggestion network to generate anchor points, dividing grids by taking the anchor points as centers, traversing the characteristic images, and then splicing all traversal results to obtain class parameters, class logic parameters and boundary frame parameters; inhibiting the class parameters, the class logic parameters and the boundary frame parameters through non-maximum values to obtain an interest region;
step 3.3.1, inputting the position information (workpiece type, labeling frame, and the like) of the interest area and the sample workpiece image into a target detection layer of the area suggestion network for target detection, and acquiring an initial workpiece type, a boundary frame and a segmentation mask, as shown in fig. 5. The structure of the area suggestion network is shown in fig. 6, and a target area which may exist is searched by scanning an image using a sliding window. The rectangles distributed over the RPN scanned area are called anchor points, as shown in fig. 7. There are nearly 20 million rectangular anchor points of different sizes and aspect ratios on the image that intersect as much as possible across the image. The RPN creates a sliding window through the convolution process and does not scan the image directly, but rather scans the backbone feature map, which allows the RPN to effectively reuse the extracted features and avoid duplicate computations, and the final anchor point will be passed on to the next stage.
And 3.3.2, combining the workpiece category, the boundary frame, the segmentation mask and the feature image, aligning the interest region, constructing a feature pyramid network with strong semantic information on all scales on the generated candidate region feature image by using a bilinear interpolation method, fusing shallow layer information such as edges and corners and deep layer information such as features and defects through a positioning network and a segmentation network of the feature pyramid network, and acquiring the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected. As shown in fig. 8, the feature pyramid network (also called a feature extractor) is composed of two paths from bottom to top and from top to bottom, and by generating a multi-layer feature mapping (multi-scale feature mapping), the quality of the obtained information is superior to that of the conventional feature detection pyramid.
And 3.4, inputting the workpiece category, the workpiece boundary box and the workpiece segmentation mask into an ROI classifier, obtaining an initial defect category and a defect boundary box as shown in FIG. 9, and then obtaining a final mask result through a segmentation mask network to finish detection.
Step 3.4.1, firstly, extracting a workpiece feature map by utilizing the information of the workpiece category, the workpiece boundary frame and the workpiece segmentation mask, inputting the workpiece feature map into an ROI pooling layer after deep convolution processing, and mapping the region selected by the workpiece image boundary frame to be detected onto the convolution feature map through the ROI pooling layer; then through the two fully connected layers to the output layer. And finally, respectively obtaining a classification maximum value and a frame regression from the output layer to obtain an initial defect class and a defect boundary frame. If the image of the workpiece to be detected has no defect, the output layer does not output the classification maximum value and the frame regression.
Step 3.4.2, the process of dividing mask network mask is shown in fig. 10, an initial defect type and a defect boundary box are used as input, then the input is divided into two branches, the upper branch performs convolution on a 7x7x256 feature map to obtain the defect type and the boundary box, the lower branch performs convolution on a 14x14x256 feature map twice, then a deconvolution operation is used for generating a 28x 28x 256 feature map, an image is enlarged and pixel points are added to improve the resolution, a 28x 28x 80 full-connection network is connected to obtain a low-resolution division mask to calculate a loss function, and finally the predicted mask is amplified to the size of the boundary box to give a final mask result.
A workpiece detection apparatus based on a deep learning network, as shown in fig. 11, includes:
the positioning module is used for acquiring a sample workpiece image, labeling the sample workpiece image and obtaining the position information of the sample workpiece image;
the identification module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region or not;
the detection module is used for detecting the defects of the workpiece image to be detected if the workpiece exists on the workpiece image to be detected; and if the workpiece does not exist on the image of the workpiece to be detected, stopping detection.
The positioning module includes:
the first acquisition module is used for acquiring a sample workpiece image;
and the marking module is used for marking the sample workpiece image and acquiring the position information of the sample workpiece image.
The identification module comprises:
the second acquisition module is used for acquiring an image of the workpiece to be detected;
and the judging module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region.
The detection module includes:
the first acquisition module is used for inputting the workpiece image to be detected into a residual error network to acquire a characteristic image;
the second acquisition module is used for inputting the characteristic image into the area suggestion network to obtain class parameters, class logic parameters and boundary frame parameters, and obtaining an interest area after the class parameters, the class logic parameters and the boundary frame parameters are inhibited by a non-maximum value;
the third acquisition module is used for inputting the position information of the interest area and the sample workpiece image into a target detection layer of a region suggestion network for target detection to obtain an initial workpiece category, a boundary frame and a segmentation mask, aligning the interest area after combining the workpiece category, the boundary frame, the segmentation mask and the feature image to obtain a feature pyramid network, and acquiring the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected through a positioning network and a segmentation network of the feature pyramid network;
and the fourth acquisition module is used for inputting the accurate workpiece category, the workpiece boundary box and the workpiece segmentation mask into the ROI classifier to obtain an initial defect category and a rough defect boundary box, and then acquiring a final mask result through a segmentation mask network.
Through the mode, the workpiece detection method can realize the classification and defect detection of the workpieces at the same time, and is suitable for flexible detection of workpieces of various types, small batches and frequent product updating.
Claims (2)
1. A workpiece detection method based on a deep learning network is characterized by comprising the following steps:
step 1, collecting a sample workpiece image, labeling the sample workpiece image, and acquiring position information of the sample workpiece image;
step 2, collecting a workpiece image to be detected, sending the workpiece image to be detected into an image cache region, and judging whether a workpiece exists in the image cache region or not;
step 3, if the workpiece exists on the workpiece image to be detected, carrying out defect detection on the workpiece image to be detected; if no workpiece exists on the image of the workpiece to be detected, stopping detection;
the step 1 specifically comprises the following steps: acquiring a sample workpiece image, and marking the sample workpiece image by an image marking tool to obtain the workpiece category and the pixel point coordinates of the sample workpiece image;
the step 2 specifically comprises the following steps: sending a workpiece image to be detected into an image cache region, drawing a suspicious region suggestion frame on the workpiece image to be detected through a region suggestion network, sending the workpiece image to be detected into a characteristic pyramid for image segmentation to obtain a suspicious region of the workpiece image to be detected, comparing the suspicious region of the workpiece image to be detected with a sample workpiece image, and judging whether a workpiece exists in the image cache region or not;
the step 3 specifically comprises the following steps:
step 3.1, inputting the workpiece image to be detected into a residual error network to obtain a characteristic image;
step 3.2, inputting the characteristic image into a region suggestion network to obtain class parameters, class logic parameters and boundary frame parameters, and obtaining an interest region after the class parameters, the class logic parameters and the boundary frame parameters are inhibited by a non-maximum value;
step 3.3, inputting the position information of the interest region and the sample workpiece image into a target detection layer of a region suggestion network for target detection to obtain an initial workpiece category, a boundary frame and a segmentation mask, aligning the interest region after combining the workpiece category, the boundary frame, the segmentation mask and the feature image to obtain a feature pyramid network, and obtaining the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected through a positioning network and a segmentation network of the feature pyramid network;
and 3.4, inputting the accurate workpiece category, the workpiece boundary box and the workpiece segmentation mask into an ROI classifier to obtain an initial defect category and a rough defect boundary box, and then obtaining a final mask result through a segmentation mask network.
2. A workpiece detection device based on a deep learning network is characterized by comprising:
the positioning module is used for acquiring a sample workpiece image, labeling the sample workpiece image and obtaining the position information of the sample workpiece image;
the identification module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region or not;
the detection module is used for detecting the defects of the workpiece image to be detected if the workpiece exists on the workpiece image to be detected; if no workpiece exists on the image of the workpiece to be detected, stopping detection;
the positioning module includes:
the first acquisition module is used for acquiring a sample workpiece image;
the marking module is used for marking the sample workpiece image and acquiring the position information of the sample workpiece image;
the identification module comprises:
the second acquisition module is used for acquiring an image of the workpiece to be detected;
the judging module is used for sending the image of the workpiece to be detected into the image cache region and judging whether the workpiece exists in the image cache region or not;
the detection module includes:
the first acquisition module is used for inputting the workpiece image to be detected into a residual error network to acquire a characteristic image;
the second acquisition module is used for inputting the characteristic image into the area suggestion network to obtain class parameters, class logic parameters and boundary frame parameters, and obtaining an interest area after the class parameters, the class logic parameters and the boundary frame parameters are inhibited by a non-maximum value;
the third acquisition module is used for inputting the position information of the interest area and the sample workpiece image into a target detection layer of a region suggestion network for target detection to obtain an initial workpiece category, a boundary frame and a segmentation mask, aligning the interest area after combining the workpiece category, the boundary frame, the segmentation mask and the feature image to obtain a feature pyramid network, and obtaining the workpiece category, the workpiece boundary frame and the workpiece segmentation mask of the workpiece image to be detected through a positioning network and a segmentation network of the feature pyramid network;
and the fourth acquisition module is used for inputting the accurate workpiece category, the workpiece boundary box and the workpiece segmentation mask into the ROI classifier to obtain an initial defect category and a rough defect boundary box, and then acquiring a final mask result through a segmentation mask network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910712712.9A CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910712712.9A CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110443791A CN110443791A (en) | 2019-11-12 |
| CN110443791B true CN110443791B (en) | 2023-04-07 |
Family
ID=68433073
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910712712.9A Active CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110443791B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111178424A (en) * | 2019-12-26 | 2020-05-19 | 中国石油大学(北京) | Petrochemical production site safety compliance real-time detection system and method |
| CN111402255B (en) * | 2020-04-10 | 2022-09-06 | 山东大学 | Side die failure detection method and system based on deep learning |
| CN111523540A (en) * | 2020-04-17 | 2020-08-11 | 中南大学 | Metal surface defect detection method based on deep learning |
| CN111951232A (en) * | 2020-07-24 | 2020-11-17 | 上海微亿智造科技有限公司 | Metal powder injection molding appearance defect detection method and system |
| CN112862770B (en) * | 2021-01-29 | 2023-02-14 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
| CN113467914A (en) * | 2021-07-08 | 2021-10-01 | 征图新视(江苏)科技股份有限公司 | Method for improving processing efficiency of image defect detection software |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106053479A (en) * | 2016-07-21 | 2016-10-26 | 湘潭大学 | System for visually detecting workpiece appearance defects based on image processing |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
| CN108144873A (en) * | 2017-12-29 | 2018-06-12 | 深圳市智能机器人研究院 | A kind of vision detection system and method |
| CN109035232A (en) * | 2018-07-24 | 2018-12-18 | 桂林电子科技大学 | A kind of intermediate plate defect inspection method merged based on deep learning and machine vision |
| CN109063708A (en) * | 2018-06-21 | 2018-12-21 | 歌尔股份有限公司 | Industrial picture characteristic recognition method and system based on contours extract |
| CN109509172A (en) * | 2018-09-25 | 2019-03-22 | 无锡动视宫原科技有限公司 | A kind of liquid crystal display flaw detection method and system based on deep learning |
| CN109859163A (en) * | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
| CN109900711A (en) * | 2019-04-02 | 2019-06-18 | 天津工业大学 | Workpiece, defect detection method based on machine vision |
-
2019
- 2019-08-02 CN CN201910712712.9A patent/CN110443791B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN106053479A (en) * | 2016-07-21 | 2016-10-26 | 湘潭大学 | System for visually detecting workpiece appearance defects based on image processing |
| CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
| CN108144873A (en) * | 2017-12-29 | 2018-06-12 | 深圳市智能机器人研究院 | A kind of vision detection system and method |
| CN109063708A (en) * | 2018-06-21 | 2018-12-21 | 歌尔股份有限公司 | Industrial picture characteristic recognition method and system based on contours extract |
| CN109035232A (en) * | 2018-07-24 | 2018-12-18 | 桂林电子科技大学 | A kind of intermediate plate defect inspection method merged based on deep learning and machine vision |
| CN109509172A (en) * | 2018-09-25 | 2019-03-22 | 无锡动视宫原科技有限公司 | A kind of liquid crystal display flaw detection method and system based on deep learning |
| CN109859163A (en) * | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
| CN109900711A (en) * | 2019-04-02 | 2019-06-18 | 天津工业大学 | Workpiece, defect detection method based on machine vision |
Non-Patent Citations (5)
| Title |
|---|
| Feature Pyramid and Hierarchical Boosting Network for Pavement Crack Detection;Fan Yang等;《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》;20190422;第21卷(第4期);第1525-1535页 * |
| Strip Steel Defect Detection Based on Saliency Map Construction Using Gaussian Pyramid Decomposition;Shengqi GUAN;《ISIJ International》;20151231;第55卷(第9期);第1950-1955页 * |
| 基于Faster R-CNN的工业CT图像缺陷检测研究;常海涛;《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》;20190115;第B022-33页 * |
| 基于深度学习的工件微小缺陷诊断方法研究;李欣遥;《中国优秀硕士学位论文全文数据库 信息科技辑》;20190715;第I138-1310页 * |
| 深度学习及其在航空发动机缺陷检测中的应用研究;旷可嘉;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20180515;第C031-107页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110443791A (en) | 2019-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110443791B (en) | Workpiece detection method and device based on deep learning network | |
| CN107543828B (en) | Workpiece surface defect detection method and system | |
| CN110175982B (en) | Defect detection method based on target detection | |
| CN106952250B (en) | Metal plate strip surface defect detection method and device based on fast R-CNN network | |
| CN114140679B (en) | Defect fusion method, device, recognition system and storage medium | |
| CN111080622B (en) | Neural network training method, workpiece surface defect classification and detection method and device | |
| CN102496161B (en) | Method for extracting contour of image of printed circuit board (PCB) | |
| CN107230203B (en) | Casting defect identification method based on human eye visual attention mechanism | |
| CN108765389A (en) | A kind of microcosmic wafer surface defects image detecting method | |
| CN111062915A (en) | Real-time steel pipe defect detection method based on improved YOLOv3 model | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN109767445B (en) | High-precision PCB defect intelligent detection method | |
| CN108665458A (en) | Transparent body surface defect is extracted and recognition methods | |
| CN110390677B (en) | Defect positioning method and system based on sliding self-matching | |
| CN112734761B (en) | Industrial product image boundary contour extraction method | |
| CN115791822A (en) | Visual detection algorithm and detection system for wafer surface defects | |
| CN112037219A (en) | Metal surface defect detection method based on two-stage convolution neural network | |
| CN111127417B (en) | Printing defect detection method based on SIFT feature matching and SSD algorithm improvement | |
| JPH03257304A (en) | Position recognizing apparatus | |
| CN109752392A (en) | A kind of pcb board defect type detection system and method | |
| CN114910480A (en) | Wafer surface defect detection method based on machine vision | |
| CN113393426A (en) | Method for detecting surface defects of rolled steel plate | |
| CN115601355A (en) | Method and device for detecting and classifying product surface defects and storage medium | |
| CN111415330A (en) | Copper foil appearance defect detection method based on deep learning | |
| CN113705564B (en) | Pointer type instrument identification reading method |
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 |