CN111784779A - Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition - Google Patents
Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition Download PDFInfo
- Publication number
- CN111784779A CN111784779A CN202010544291.6A CN202010544291A CN111784779A CN 111784779 A CN111784779 A CN 111784779A CN 202010544291 A CN202010544291 A CN 202010544291A CN 111784779 A CN111784779 A CN 111784779A
- Authority
- CN
- China
- Prior art keywords
- checkerboard
- image
- points
- calibration
- calibration plate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The invention discloses a chessboard pattern image recognition and positioning system and method based on convolutional neural network and nested contour recognition, wherein the system consists of a chessboard pattern calibration plate, an image acquisition system, an illumination system, a feature extraction system and a contour recognition system; asymmetric circular feature points are arranged on the edge of the calibration plate image of the checkerboard calibration plate; the image acquisition system is used for acquiring checkerboard calibration board images; the lighting system is used for lighting the checkerboard calibration board; the characteristic extraction system is used for extracting circular characteristic points at the edge of the calibration plate, and storing image coordinates of the characteristic points and images after the characteristic points are extracted; the contour recognition system is used for extracting the checkerboard points after the background is removed, and storing the total number of the recognized calibration board image grid points, the image coordinates and the space position coordinates of the checkerboard points. The invention adopts the self-designed checkerboard calibration plate, can be effectively applied to multi-camera calibration and rotation calibration, and improves the accuracy of checkerboard point coordinate extraction.
Description
Technical Field
The invention relates to a camera calibration technology, in particular to a checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition.
Background
Machine vision has long been used in industrial automation systems to improve production quality and yield by replacing traditional manual inspection, and its application in industrial production, intelligent transportation, security monitoring and other fields is accepted by users. The machine vision system comprises two major links of image acquisition and image processing, is composed of a light source, a lens, an industrial camera, image processing software and other core components, and mainly has three functions of positioning, identification and detection. In the whole imaging process of the machine vision system, high-precision system calibration is the basis and key point for realizing high-definition imaging and has direct influence on final application, so that camera calibration is a crucial link in the production of machine vision application, and the accuracy of feature point extraction directly influences the accuracy of camera calibration.
The traditional camera calibration directly adopts a common checkerboard to calibrate the camera, so that when the checkerboard rotates or multiple cameras are needed to calibrate, the grid point positioning sequence of the checkerboard is inconsistent or the positioning is inaccurate due to background interference.
Disclosure of Invention
The invention aims to provide a checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition, and accuracy and reliability of camera calibration are improved.
The technical solution for realizing the purpose of the invention is as follows: a chessboard pattern image recognition and positioning system based on convolutional neural network and nested contour recognition is composed of a chessboard pattern calibration plate, an image acquisition system, an illumination system, a feature extraction system and a contour recognition system;
asymmetric circular feature points are arranged on the edge of the calibration plate image of the checkerboard calibration plate;
the image acquisition system is used for acquiring checkerboard calibration board images;
the illumination system is composed of a machine vision area light source and is used for illuminating the chessboard pattern calibration plate;
the characteristic extraction system is used for extracting circular characteristic points at the edge of the calibration plate, and storing image coordinates of the characteristic points and images after the characteristic points are extracted;
the contour recognition system is used for extracting the checkerboard points after the background is removed, and storing the total number of the recognized calibration board image grid points, the image coordinates and the space position coordinates of the checkerboard points.
Furthermore, the checkerboard calibration plate is an asymmetric checkerboard calibration pattern with 5 feature points at corners, each check in the checkerboard is a square black-white checkerboard, the corners of the calibration plate adopt circular feature points as the feature points of the calibration plate, and one corner has two circular feature points.
Furthermore, a support rod is connected with the rotating table below the chessboard pattern calibration plate.
The invention also provides a chessboard pattern image identification and positioning method based on convolutional neural network and nested contour identification, which comprises the following steps:
collecting a chessboard pattern on a chessboard pattern calibration plate, wherein asymmetric circular feature points are arranged on the edge of the chessboard pattern calibration pattern;
extracting circular characteristic points at the edge of the calibration plate, and storing image coordinates of the characteristic points and an image after the characteristic points are extracted;
and extracting the checkerboard points after the background is removed, and storing the total points of the identified image grid points of the calibration board, the image coordinates and the space position coordinates of the checkerboard points.
Furthermore, the checkerboard calibration plate is an asymmetric checkerboard calibration pattern with 5 feature points at corners, each check in the checkerboard is a square black-white checkerboard, the corners of the calibration plate adopt circular feature points as the feature points of the calibration plate, and one corner has two circular feature points.
Furthermore, a support rod is connected with the rotating table below the chessboard pattern calibration plate.
Further, extracting circular characteristic points of the edge of the calibration plate, specifically comprising the following steps: establishing an anchor point frame feature identification network, firstly marking circular feature points in a calibration plate by using a rectangular frame by using the network, establishing a 1x1 frame and a 4x4 anchor point frame for each pixel point of an image in the training process, judging the overlapping area of the anchor point frame and the rectangular frame marked in the training set, and judging the anchor point frame to be effective if the overlapping area is more than fifty percent; and training to obtain a convolutional neural network capable of automatically identifying the feature points.
Further, after extracting all the feature point coordinates, removing a background from the chessboard pattern calibration plate image, performing binarization threshold calculation, performing binarization on the calibration plate image, and storing the binarized calibration plate image; and finally, identifying the grid points of the chessboard by using the findChessboardCorrers () function of the existing opencv, extracting the coordinates of all the grid points in the calibration board image, if the number of the found grid points is the same as the number of the preset grid points, sequentially identifying and storing the coordinates of all the grid points of the chessboard according to the Z-shaped sequence, and finally remapping the coordinates back to the original image.
Compared with the prior art, the invention has the following remarkable advantages: (1) the chessboard marking board containing the feature points is designed independently, and the asymmetric feature points are arranged on the edge of the image of the marking board, so that the chessboard marking board can be identified in a uniform sequence and direction no matter how the marking board rotates or turns over, the chessboard marking board can be effectively applied to multi-camera marking and rotation marking, and the marking accuracy is improved; (2) the characteristic extraction system is used for extracting circular characteristic points at the edge of the calibration plate, removing the influence of the background on the extraction of the checkerboard points, improving the extraction accuracy of the checkerboard points, and storing the image coordinates of the characteristic points and the image after the characteristic points are extracted so as to check and verify the identification accuracy; (3) the contour recognition system extracts the checkerboard points after the background is removed, and stores the total number of the recognized calibration board image grid points, the image coordinates and the space position coordinates of the checkerboard points.
Drawings
Fig. 1 is an overall configuration diagram of a checkerboard image recognition positioning system.
FIG. 2 is a checkerboard calibration plate pattern.
Fig. 3 is a schematic structural diagram of a convolutional neural network for feature point extraction.
Fig. 4 is an image in which the feature point coordinates of the calibration plate are extracted.
Fig. 5 is an image binarized based on the extracted feature point coordinates.
Fig. 6 is a calibration plate image with background removed.
Fig. 7 is an output image after nested contour recognition.
Detailed Description
As shown in FIG. 1, the invention provides a checkerboard image recognition and positioning system based on convolutional neural network and nested contour recognition, which comprises a checkerboard calibration plate, an image acquisition system, an illumination system, a feature extraction system and a contour recognition system.
The invention designs an asymmetric checkerboard calibration pattern with 5 characteristic points at corners, as shown in fig. 2, each grid in the checkerboard calibration plate is a square black-white checkerboard, the number of the checkerboard points is 8x8, and the side length of each grid is 10 mm. Compared with the traditional chessboard calibration plate, the chessboard calibration plate provided by the invention is added with 5 circular feature points, so that the arrangement sequence of the fixed points of the calibration plate can be ensured to be always consistent when the calibration plate rotates and is measured for many times, and the inaccuracy of calibration results caused by inconsistent grid point distances of the calibration plate is avoided. In addition, 5 feature points are added to be used as grid point area limitation, and therefore the identification error of the calibration point caused by environmental factors can be effectively removed. The round feature points are adopted by the corners of the calibration board as the feature points of the calibration board, so that the condition that the feature points are used as checkerboard points by a grid point identification algorithm to be identified wrongly can be effectively avoided. The feature points of the four corners are 100mm farthest apart, with one corner having two rounded feature points. The 8x8 lattice points are adopted, the side length of the lattice points is 10mm, and the advantage is that the size of each lattice point in the camera view field is proper, and the number of the lattice points is not too dense or sparse; the method can meet the calibration requirements of most industrial cameras, and can also ensure the accuracy and precision of checkerboard point identification. In the using process of the system, firstly, images of a black and white chessboard pattern calibration plate are collected through a high-definition camera, after the collected images are input into a trained convolutional neural network system, the convolutional neural network can calculate the image coordinates of the identified circular feature points on the calibration plate, and the coordinates of the 5 feature points are extracted. The environment background is removed through the image coordinates of the 5 characteristic points, the influence of environmental factors on the extraction of the checkerboard calibration checkerboard points is reduced, and the accuracy and the reliability of recognition are improved. Because the numbers of the circular characteristic points of the four corners of the calibration plate are different, the corner with two circular characteristic points is used as the third corner of the whole calibration plate in the clockwise direction, and therefore the identification sequence of the grid points of the chessboard grid calibration plate is ensured to be consistent no matter how the calibration plate rotates. After the convolutional neural network extracts the characteristic points and removes the environmental background, the grid points of the chessboard grid calibration plate are automatically identified through nested contour identification, all the grid points of the chessboard are extracted, and grid point coordinates are output according to the sequence from the first corner to the back according to the Z-shaped sequence.
The chess board check calibration board material is aluminum plate, prints the calibration board pattern to aluminum surface to designed supporting according to the shell, there are M6 and M4's screw in shell both sides, make things convenient for the installation and the fixed of calibration board, the below adopts branch and revolving stage to be connected, is convenient for gather the calibration board image of different angles. Because only one corner in the checkerboard calibration plate contains two circular characteristic points, the collected checkerboard points can be kept consistent in arrangement sequence no matter how the calibration plate rotates and turns.
The image acquisition system adopts a high-definition camera, the camera lens adopts an industrial lens with the focal length of 12mm, the acquired image is a high-definition image with the resolution of 1292x964, the camera performs data transmission with a computer through a serial port, and the data is finally stored in the computer in a bmp format.
As shown in fig. 3, the feature extraction system establishes a training set containing 624 calibration plate images, and marks circular feature points in the calibration plate with rectangular frames, establishes a 1x1 frame and an anchor point frame of 4x4 for each pixel point of each image in the training process, and determines the overlapping area of the anchor point frame and the rectangular frame marked in the training set, and if the overlapping area is greater than fifty percent, determines that the anchor point frame is valid; the convolutional neural network capable of automatically identifying the feature points is obtained through 40000 times of training, and the coordinates of the 5 circular feature points on the collected calibration plate image are extracted. After the convolutional neural network is trained by a VGG-64 convolutional neural network after an image is input into a system, the network is divided into an upper line and a lower line for training, one line trains the coordinate position (x, y) of an extracted anchor point and the length (w, h) of an anchor point frame, one line trains whether the content in the anchor point frame is reliable to the marked feature or not, and finally the coordinate position and the size of the anchor point frame similar to the marked feature in the image, namely the coordinates of 5 feature points in the extracted checkerboard are judged by combining the two training results, the coordinates of the 5 feature points and the recognition accuracy are output and stored, after the recognition accuracy of the feature points obtained by a feature recognition system is compared with a set threshold, if the recognized accuracy is greater than the set threshold, the recognition is judged to be accurate, and if the recognized accuracy is less than the set threshold, and manually marking the feature points by adopting a manual feature point marking mode, and then outputting the image coordinates of each feature point according to the sequence from the first corner point to the fourth corner point.
The contour recognition system eliminates the environmental background in the collected checkerboard image through the nested contour recognition system according to the circular feature point coordinates extracted by the feature extraction system, and reduces the influence of environmental factors on the extraction of the checkerboard points. And then, performing binarization on the image by taking the center of the image as a threshold value, judging whether the areas of black and white areas in the area are the same, if so, performing binarization on the image according to the threshold value, and if the difference is more than ten percent of the total area, adjusting the threshold value by a dichotomy until the areas of the black and white areas are basically the same. And recognizing the grid points by using an opencv grid point recognition algorithm to extract image coordinates of the checkerboard grid points, and re-projecting the image coordinates back to the original image in a Z shape according to the sequence from the first corner to the fourth corner of the calibration plate for arrangement.
Compared with the traditional calibration plate characteristic extraction, the calibration plate image with more advantages is designed, so that the checkerboard point arrangement sequence on the calibration plate can be always kept consistent when the multi-camera calibration and the calibration plate rotate and turn. The automatic identification of the feature points is realized through the convolutional neural network, compared with the traditional opencv algorithm, the fuzzy image features can be identified, the practicability of feature extraction and the accuracy of feature identification are improved, the environment background is eliminated by means of the extracted feature point coordinates, and the influence of environment factors on checkerboard point extraction is reduced. The accuracy of checkerboard contour recognition is greatly improved.
The present invention will be described in detail with reference to examples.
Examples
The identification positioning system is shown in fig. 1, and the calibration plate pattern is shown in fig. 2, in this embodiment, an aluminum calibration plate 100 is first installed in an aluminum fixing clamp, a support rod and a rotating table 101 are installed, and the rotating table is fixed on a shockproof platform to serve as a measured calibration plate system; installing the camera 200 on the bracket 201 and connecting the camera with a computer for collecting images through a serial port line, turning on the machine vision light source 300, adjusting the illumination uniformity and adjusting the focal length of a camera lens until the camera can collect complete and clear checkerboard calibration plate images; after collecting the calibration board image, inputting the image in the computer into the feature recognition system 400, the feature recognition system automatically extracts the coordinates of 5 feature points in the chessboard calibration board through the trained convolutional neural network, and outputs and stores the coordinates and recognition accuracy of the 5 feature points, after comparing the recognition accuracy of the feature points obtained by the feature recognition system with the set threshold, if the recognized feature accuracy is greater than the set threshold, the recognition is determined to be accurate, if the recognized feature accuracy is less than the set threshold, the feature points are manually marked in a manner of manually marking the feature points, and then the image coordinates of each feature point are output according to the sequence from the first edge point to the fourth edge point.
The output results are shown in the following table:
| 5 |
| 461.250092,243.855927 |
| 933.780518,214.238922 |
| 963.284302,688.361572 |
| 900.955017,691.970581 |
| 486.356750,715.283264 |
wherein 5 in the first row is the number of the identified feature points, the second row is the coordinate of the first feature point, the second row is the coordinate of the second feature point, the third row is the coordinate of the third feature point, the fourth row is the coordinate of the fourth point, and the fifth row is the coordinate of the fifth feature point.
After extracting all the feature point coordinates, the feature point coordinates and the image are transmitted to the nested contour recognition system 401, the background of the chessboard pattern calibration plate image is removed according to the image coordinates of the feature points, the binarization threshold value calculation is carried out, the binarization is carried out on the calibration plate image, and the binarized calibration plate image is stored.
Fig. 4 is an image obtained after extracting the coordinates of the feature points of the calibration plate, and the image is a blurred image with an inverted tilt, where a point frame is an identified feature frame, and the above numerical value is the reliability of feature point identification. Fig. 5 is an image binarized based on the extracted feature point coordinates. Fig. 6 is a calibration plate image with background removed.
And finally, extracting the coordinates of all grid points in the calibration board image by a checkerboard grid point recognition algorithm, sequentially recognizing and storing the coordinates of all grid points of the checkerboard according to the sequence of the zigzag, and finally remapping the coordinates back to the original image, wherein the result is shown in fig. 7. Wherein Point is 5 identified feature points, (0, 0), (1, 0), (2, 0), (3, 0) are 4 identified points of the calibration board, and the middle is identified checkerboard points arranged in a zigzag manner.
The data storage results are as follows:
| 64 |
| 562.839417,368.580475 |
| 0.000000,0.000000,0.000000 |
| 598.800232,363.070679 |
| 10.000000,0.000000,0.000000 |
| … |
64 in the first row is the number of the total points identified, and the data in the second row is the image coordinates of the first points in the collected checkerboard image; the third row is the 3-dimensional world coordinate of the point, and the 3 values respectively represent coordinate values in the x, y and z directions; the fourth row is the image coordinates of the second point, the fifth row is the 3-dimensional coordinates of the point, and so on for all subsequent grid points, for a total of 64 grid point coordinates.
Claims (8)
1. A chessboard pattern image recognition and positioning system based on convolutional neural network and nested contour recognition is characterized by comprising a chessboard pattern calibration plate, an image acquisition system, an illumination system, a feature extraction system and a contour recognition system;
asymmetric circular feature points are arranged on the edge of the calibration plate image of the checkerboard calibration plate;
the image acquisition system is used for acquiring checkerboard calibration board images;
the illumination system is composed of a machine vision area light source and is used for illuminating the chessboard pattern calibration plate;
the characteristic extraction system is used for extracting circular characteristic points at the edge of the calibration plate, and storing image coordinates of the characteristic points and images after the characteristic points are extracted;
the contour recognition system is used for extracting the checkerboard points after the background is removed, and storing the total number of the recognized calibration board image grid points, the image coordinates and the space position coordinates of the checkerboard points.
2. A checkerboard image recognition and positioning system as claimed in claim 1, wherein said checkerboard is an asymmetric checkerboard pattern with 5 feature points at the corners, each cell in the checkerboard is a square black and white checkerboard, the corners of the checkerboard use circular feature points as the feature points of the checkerboard, and one corner has two circular feature points.
3. A checkerboard image recognition and localization system as claimed in claim 1 or 2, wherein said checkerboard calibration plate is connected to the rotation stage with struts underneath.
4. A chessboard pattern image recognition and positioning method based on convolutional neural network and nested contour recognition is characterized by comprising the following steps:
collecting a chessboard pattern on a chessboard pattern calibration plate, wherein asymmetric circular feature points are arranged on the edge of the chessboard pattern calibration pattern;
extracting circular characteristic points at the edge of the calibration plate, and storing image coordinates of the characteristic points and an image after the characteristic points are extracted;
and extracting the checkerboard points after the background is removed, and storing the total points of the identified image grid points of the calibration board, the image coordinates and the space position coordinates of the checkerboard points.
5. A checkerboard image recognition and positioning method based on convolutional neural network and nested profile recognition as claimed in claim 4, wherein said checkerboard calibration is an asymmetric checkerboard calibration pattern with 5 feature points at the corners, each cell in the checkerboard is a square black and white checkerboard, the calibration board corners use circular feature points as the feature points of the calibration board, and one corner has two circular feature points.
6. A chessboard image recognition and positioning method based on convolutional neural network and nested contour recognition as claimed in claim 4, wherein the chessboard pattern calibration plate is connected with a rotating platform by using struts below it.
7. The chessboard pattern image recognition and positioning method based on convolutional neural network and nested contour recognition as claimed in claim 4, wherein the circular feature points of the edge of the calibration plate are extracted by the following specific method: establishing an anchor point frame feature identification network, firstly marking circular feature points in a calibration plate by using rectangular frames, establishing a 1x1 frame and a 4x4 anchor point frame for each pixel point of an image in the training process, judging the overlapping area of the anchor point frames and the rectangular frames marked in the training set, and judging the anchor point frames to be effective if the overlapping area is more than fifty percent; and training to obtain a convolutional neural network capable of automatically identifying the feature points.
8. The chessboard pattern image recognition and positioning method based on convolutional neural network and nested contour recognition as claimed in claim 7, wherein after all feature point coordinates are extracted, the background of the chessboard pattern calibration plate image is removed, binarization threshold calculation is performed, then binarization is performed on the calibration plate image, and the binarized calibration plate image is saved; and finally, identifying the grid points of the chessboard by using the findChessboardCorrers () function of the existing opencv, extracting the coordinates of all the grid points in the calibration board image, if the number of the found grid points is the same as the number of the preset grid points, sequentially identifying and storing the coordinates of all the grid points of the chessboard according to the Z-shaped sequence, and finally remapping the coordinates back to the original image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010544291.6A CN111784779A (en) | 2020-06-15 | 2020-06-15 | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010544291.6A CN111784779A (en) | 2020-06-15 | 2020-06-15 | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111784779A true CN111784779A (en) | 2020-10-16 |
Family
ID=72756639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010544291.6A Pending CN111784779A (en) | 2020-06-15 | 2020-06-15 | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111784779A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113916892A (en) * | 2021-09-28 | 2022-01-11 | 南京理工大学 | Multi-view vision-based brake disc gluing defect detection device and method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103942796A (en) * | 2014-04-23 | 2014-07-23 | 清华大学 | High-precision projector and camera calibration system and method |
| CN110660107A (en) * | 2019-08-23 | 2020-01-07 | 贝壳技术有限公司 | Plane calibration plate, calibration data acquisition method and system |
| CN110689579A (en) * | 2019-10-18 | 2020-01-14 | 华中科技大学 | Rapid monocular vision pose measurement method and measurement system based on cooperative target |
-
2020
- 2020-06-15 CN CN202010544291.6A patent/CN111784779A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103942796A (en) * | 2014-04-23 | 2014-07-23 | 清华大学 | High-precision projector and camera calibration system and method |
| CN110660107A (en) * | 2019-08-23 | 2020-01-07 | 贝壳技术有限公司 | Plane calibration plate, calibration data acquisition method and system |
| CN110689579A (en) * | 2019-10-18 | 2020-01-14 | 华中科技大学 | Rapid monocular vision pose measurement method and measurement system based on cooperative target |
Non-Patent Citations (1)
| Title |
|---|
| 韩鑫: "摄像头模组组装及成像质量一站式检测方法的研究", 《中国优秀硕士论文全文数据库》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113916892A (en) * | 2021-09-28 | 2022-01-11 | 南京理工大学 | Multi-view vision-based brake disc gluing defect detection device and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111242938B (en) | Automatic checking method for state of pressing plate | |
| CN108562250B (en) | Keyboard keycap flatness rapid measurement method and device based on structured light imaging | |
| CN111260788B (en) | Power distribution cabinet switch state identification method based on binocular vision | |
| CN111284154B (en) | Seal control machine seal control method, device and system based on image recognition | |
| CN102704215B (en) | Automatic cutting method of embroidery cloth based on combination of DST file parsing and machine vision | |
| CN111308448A (en) | Image acquisition equipment and radar external parameter determination method and device | |
| CN110033407B (en) | Shield tunnel surface image calibration method, splicing method and splicing system | |
| CN104700395A (en) | Method and system for detecting appearance crack of structure | |
| CN107092905B (en) | Method for positioning instrument to be identified of power inspection robot | |
| CN113608378B (en) | Full-automatic defect detection method and system based on LCD (liquid crystal display) process | |
| CN109406527B (en) | System and method for detecting fine appearance defects of micro camera module lens | |
| CN108709500B (en) | Circuit board element positioning and matching method | |
| CN110807763A (en) | Method and system for detecting ceramic tile surface bulge | |
| CN110599404A (en) | Circuit board microscopic image splicing method and device and information data processing terminal | |
| CN111784779A (en) | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition | |
| CN111242939B (en) | Method for identifying state of pressing plate | |
| CN110596118A (en) | Print pattern detection method and print pattern detection device | |
| CN105427237A (en) | Steel mesh image registration and detection method for large-format optical measurement system | |
| CN110136248B (en) | Transmission shell three-dimensional reconstruction device and method based on binocular stereoscopic vision | |
| CN111738936A (en) | Image processing-based multi-plant rice spike length measuring method | |
| CN115082504B (en) | Light spot identification method for solar photovoltaic panel | |
| CN111402343A (en) | High-precision calibration plate and calibration method | |
| CN111242869B (en) | Method for background filtering of photo | |
| CN103279953A (en) | Machine vision calibration system based on LabVIEW platform | |
| CN111260625B (en) | Automatic extraction method for offset printing large image detection area |
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 |