CN114187247A - Ampoule bottle printing character defect detection method based on image registration - Google Patents
Ampoule bottle printing character defect detection method based on image registration Download PDFInfo
- Publication number
- CN114187247A CN114187247A CN202111432045.2A CN202111432045A CN114187247A CN 114187247 A CN114187247 A CN 114187247A CN 202111432045 A CN202111432045 A CN 202111432045A CN 114187247 A CN114187247 A CN 114187247A
- Authority
- CN
- China
- Prior art keywords
- image
- registered
- ampoule bottle
- difference
- template
- 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
- 239000003708 ampul Substances 0.000 title claims abstract description 34
- 230000007547 defect Effects 0.000 title claims abstract description 32
- 238000001514 detection method Methods 0.000 title claims abstract description 32
- 239000011159 matrix material Substances 0.000 claims abstract description 19
- 230000009466 transformation Effects 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 abstract description 4
- 238000012937 correction Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 4
- 239000003814 drug Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 229940079593 drug Drugs 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- G06T5/80—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30144—Printing quality
Abstract
The invention discloses an ampoule bottle printed word defect detection method based on image registration, which comprises the following steps of 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered; step 2: extracting and describing feature points of the template image by using an SURF algorithm, and extracting and describing the feature points of the image to be registered; and step 3: matching the feature points by using a FLANN matching algorithm; and 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched feature point pairs; and 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated; step 6: and carrying out image difference on the template image and the calibrated image to be registered to obtain a difference image, and judging whether the printed word has defects according to the difference image. The invention can quickly and effectively realize the characteristic extraction, matching, correction and detection of the ampoule bottle printing digital image, and the detection process uses less time.
Description
Technical Field
The invention belongs to the field of industrial printed word defect detection, and particularly relates to an ampoule bottle printed word defect detection method based on image registration.
Background
In recent years, the quality safety problem of medicines is concerned by more and more people, and the quality detection of medicines is also enhanced in China. Aiming at the field of ampoule bottle printing, the quality requirement for printing characters is stricter, and the problem of medicine safety caused by incomplete printing characters or incomplete information of the ampoule bottle is avoided.
The printing equipment rear end that present medicine bottle manufacturing enterprise adopted does not have supporting printing check out test set, can only hire the manual work and detect ampoule printing quality, and the manual detection has the defect that can't avoid:
1. the condition of missing detection is easily caused, the eyes can be tired due to long-time work of workers, some defect information is easily missed, and the local judgment of some defects is wrong.
2. The qualified standards of the quality cannot be completely unified, and different people detecting the same product may make different judgments on the product due to subjective consciousness.
3. Slow, costly, inefficient, and require a large amount of labor and capital.
Therefore, based on the above problems, a novel ampoule bottle printed word defect detection method is urgently needed to replace the manual work and improve the production efficiency.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide an ampoule bottle printed word defect detection method based on image registration, and solves the problems in the background art.
The invention provides the following technical scheme:
an ampoule bottle printing word defect detection method based on image registration comprises the following steps:
step 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered;
step 2: extracting and describing feature points of the template image by using an SURF algorithm, and extracting and describing the feature points of the image to be registered;
and step 3: matching the feature points by using a FLANN matching algorithm;
and 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched feature point pairs;
and 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated;
step 6: and carrying out image difference on the template image and the calibrated image to be registered to obtain a difference image, and judging whether the printed word has defects according to the difference image.
Preferably, in step 1, a CMOS line industrial camera is used to capture an image, the image sensor of which has two rows of photosensitive elements.
Preferably, in step 2, the input picture is adjusted to a uniform size before the feature point extraction.
Preferably, in step 3, after feature point matching is performed by using a FLANN matching algorithm, the euclidean distance of each feature point pair is calculated, and feature point pairs with the euclidean distance greater than a set threshold are removed.
Preferably, in step 4, the RANSAC operator is used to eliminate the mismatching point pairs when calculating the transformation matrix.
Preferably, in step 6, the template image and the calibrated image to be registered are grayed, and then image difference is performed.
Preferably, after the differential image is obtained, binarization processing and filtering processing are sequentially performed on the differential image, the number of pixels with a pixel value of 255 in the differential image is counted, and if the counted number is greater than a set threshold value, a missing printing condition is considered to exist.
Compared with the prior art, the invention has the following beneficial effects:
(1) the ampoule bottle printing character defect detection method based on image registration can quickly and effectively realize the extraction, matching, correction and detection of the features of the ampoule bottle printing character image, the image features have the characteristics of scale invariance, light non-deformability and rotation invariance, and the detection process is less in time consumption.
(2) According to the ampoule bottle printed word defect detection method based on image registration, the FLANN matching algorithm is used for matching the feature points, so that the feature point matching precision is improved, and the feature matching error is reduced.
(3) According to the ampoule bottle printing character defect detection method based on image registration, images to be registered are calibrated through the transformation matrix, image distortion is eliminated, adverse effects on detection work due to image distortion are avoided, the detection result is improved, and the detection efficiency is improved.
(4) According to the ampoule bottle printing word defect detection method based on image registration, when a transformation matrix is calculated, RANSAC operators are used, mismatching point pairs are eliminated, the influence of the mismatching point pairs on the detection process is reduced, and the detection precision is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of the present invention.
Fig. 2 shows a template image and an image to be registered.
Fig. 3 is a schematic diagram of feature point pair matching after removing mismatching.
Fig. 4 is an image of an image to be registered after being subjected to transformation matrix calibration and restoration.
Fig. 5 is an effect diagram of image difference between the template image and the calibrated image to be registered.
Fig. 6 is an effect diagram of the difference image after binarization processing.
Fig. 7 is a diagram of the effect of the difference image after being filtered.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be described in detail and completely with reference to the accompanying drawings. It is to be understood that the described embodiments are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
The first embodiment is as follows:
as shown in fig. 1, an ampoule bottle printed word defect detection method based on image registration includes the following steps:
step 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered; step 2: extracting and describing feature points of the template image by using an SURF algorithm, and extracting and describing the feature points of the image to be registered; and step 3: matching the feature points by using a FLANN matching algorithm; and 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched feature point pairs; and 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated; step 6: and carrying out image difference on the template image and the calibrated image to be registered to obtain a difference image, and judging whether the printed word has defects according to the difference image. In step 1, a CMOS line industrial camera is used to capture an image, the image sensor of which has two rows of photosensitive elements. In step 2, the input picture is adjusted to a uniform size before feature point extraction. In step 3, after feature point matching is performed by using a FLANN matching algorithm, the Euclidean distance of each feature point pair is calculated, and the feature point pairs with the Euclidean distances larger than a set threshold value are removed. In step 4, when calculating the transformation matrix, using RANSAC operator to eliminate the mismatching point pairs. In step 6, firstly, graying the template image and the calibrated image to be registered, then, carrying out image difference to obtain a difference image, and then, sequentially carrying out binarization processing and filtering processing on the difference image, counting the number of pixels with a pixel value of 255 in the difference image, and if the counted number is greater than a set threshold value, determining that a missing printing condition exists.
Example two
An ampoule bottle printing word defect detection method based on image registration comprises the following steps: step 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered, wherein the template image and the image to be registered are shown in figure 2; step 2: extracting and describing feature points of the template image by using an SURF algorithm, extracting and describing the feature points of the image to be registered, and reducing the size of the input image to 250 x 200 before extracting the feature points; and step 3: using a FLANN matching algorithm to match the feature points, as shown in FIG. 3, after the feature points are matched, calculating the Euclidean distance of each feature point pair, setting the value of 2 times of the minimum Euclidean distance as a threshold value, and removing the feature point pairs with the Euclidean distance larger than the set threshold value; and 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched characteristic point pairs, and simultaneously, removing mismatching point pairs again by using an RANSAC operator, removing interference generated by the mismatching point pairs and improving the matching precision; and 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated, as shown in FIG. 4; step 6: firstly, graying the template image and the calibrated image to be registered, then carrying out image difference to obtain a difference image as shown in fig. 5, then carrying out binarization processing and filtering processing on the difference image in sequence as shown in fig. 6 and 7, then counting the number of pixel points with the pixel value of 255 in the difference image, if the counted number is greater than a set threshold value, determining that a missing printing condition exists, and if the counted number is less than the set threshold value, determining that printing is complete.
EXAMPLE III
As shown in fig. 1, an ampoule bottle printed word defect detection method based on image registration includes the following steps:
step 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered;
step 2: utilizing an SURF algorithm to extract and describe feature points of the template image, and extracting and describing the feature points of the image to be registered, wherein the description of the feature points comprises two steps of feature point direction distribution and 128-dimensional vector description, and the direction distribution of the feature points comprises the following steps: the gradient directions of all pixels in the neighborhood around the feature point are calculated by the Sift, a gradient direction histogram is generated and is normalized to be the gradient direction histogram of 0-360 degrees to 36 directions, and the direction represented by the main component of the gradient histogram is taken as the direction of the feature point; the 128-dimensional vector describes: based on the gradient direction histogram expansion, 4 × 4 blocks around the feature point are taken, 8 gradient directions are extracted from each block, and 128 directions are taken as descriptors of the features.
And step 3: matching the feature points by using a FLANN matching algorithm; the result of the feature point matching will obtain a corresponding relationship list of the two feature sets. The first set of feature sets is referred to as the training set and the second set is referred to as the query set. FLANN trains a matcher to increase the speed of matching before invoking the matching function. The training phase was to optimize the performance of cv:FlanBasedMatcher. the train class builds an index tree of feature sets. Matching each characteristic point of the query image with a train matcher to find out the best match; the method comprises the steps of matching features of query images with a trainer one by one, enabling each query image feature point to have a best match, then verifying the correctness of the match, and eliminating the match with large error by setting a truncation value.
And 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched feature point pairs; and when the transformation matrix is calculated, using an RANSAC operator to remove the mismatching point pairs. RANSAC estimates the model by repeatedly selecting a data set, and continuously iterates until a better model is estimated.
The specific implementation steps are as follows:
(1) selecting a minimum data set which can be used for estimating a model;
(2) using this data set to compute a data model;
(3) all data are brought into the model, the number of 'inner points' is calculated, and the data which are suitable for the current iteration and are in a certain error range are accumulated;
(4) comparing the number of the 'interior points' of the current model and the best model deduced before, and recording the model parameters of the maximum 'interior points' number and the 'interior points' number;
and repeating the steps 1-4 until the iteration is finished or the current model is a better model.
And 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated;
step 6: the method comprises the steps of carrying out image difference on a template image and a calibrated image to be registered to obtain a difference image, judging whether a printed word has defects according to the difference image, carrying out graying processing on the template image and the calibrated image to be registered first, then carrying out image difference to obtain the difference image, then carrying out binarization processing and filtering processing on the difference image in sequence, counting the number of pixels with the pixel value of 255 in the difference image, and if the counted number is larger than a set threshold value, determining that a missing printing condition exists.
The device obtained by the technical scheme is an ampoule bottle printing character defect detection method based on image registration, can quickly and effectively realize the processes of feature extraction, matching, correction and detection of ampoule bottle printing digital images, and has the advantages of scale invariance, light non-deformability, rotation invariance and less time consumption.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and it will be apparent to those skilled in the art that various modifications and variations can be made in the present invention; any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. An ampoule bottle printing word defect detection method based on image registration is characterized by comprising the following steps:
step 1: acquiring an ampoule bottle printing digital image by using a linear array industrial camera to obtain a template image and an image to be registered;
step 2: extracting and describing feature points of the template image by using an SURF algorithm, and extracting and describing the feature points of the image to be registered;
and step 3: matching the feature points by using a FLANN matching algorithm;
and 4, step 4: calculating a transformation matrix of the image to be registered mapped to the template image according to the matched feature point pairs;
and 5: the image to be registered is subjected to matrix transformation, calibration is carried out, and image distortion is eliminated;
step 6: and carrying out image difference on the template image and the calibrated image to be registered to obtain a difference image, and judging whether the printed word has defects according to the difference image.
2. The ampoule bottle printed word defect detection method based on image registration as claimed in claim 1, characterized in that in step 1, a CMOS line industrial camera is used to capture images, and an image sensor thereof has two lines of photosensitive elements.
3. The ampoule bottle printed word defect detection method based on image registration according to claim 1, characterized in that in step 2, the input picture is adjusted to a uniform size before feature point extraction.
4. The method for detecting defects of printed words on ampoule bottles based on image registration as claimed in claim 1, wherein in step 3, after feature point matching is performed by using a FLANN matching algorithm, the euclidean distance of each feature point pair is calculated, and feature point pairs with euclidean distances greater than a set threshold are removed.
5. The method for detecting defects of ampoule bottle printing words based on image registration as claimed in claim 1, wherein in step 4, a RANSAC operator is used to eliminate the mismatching point pairs when calculating the transformation matrix.
6. The method for detecting the defect of the printed word on the ampoule bottle based on the image registration as claimed in claim 1, wherein in step 6, the template image and the calibrated image to be registered are grayed first, and then image difference is performed.
7. The method for detecting the ampoule bottle printed word defect based on the image registration as claimed in claim 6, wherein after the difference image is obtained, the difference image is subjected to binarization processing and filtering processing in sequence, the number of pixels with a pixel value of 255 in the difference image is counted, and if the counted number is greater than a set threshold value, a missing printing condition is considered to exist.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111432045.2A CN114187247A (en) | 2021-11-29 | 2021-11-29 | Ampoule bottle printing character defect detection method based on image registration |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111432045.2A CN114187247A (en) | 2021-11-29 | 2021-11-29 | Ampoule bottle printing character defect detection method based on image registration |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114187247A true CN114187247A (en) | 2022-03-15 |
Family
ID=80541697
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111432045.2A Pending CN114187247A (en) | 2021-11-29 | 2021-11-29 | Ampoule bottle printing character defect detection method based on image registration |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114187247A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114757913A (en) * | 2022-04-15 | 2022-07-15 | 电子科技大学 | Display screen defect detection method |
WO2024087640A1 (en) * | 2022-10-26 | 2024-05-02 | 上海第二工业大学 | Printed circuit board welding spot defect detection method based on digital image processing |
-
2021
- 2021-11-29 CN CN202111432045.2A patent/CN114187247A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114757913A (en) * | 2022-04-15 | 2022-07-15 | 电子科技大学 | Display screen defect detection method |
WO2024087640A1 (en) * | 2022-10-26 | 2024-05-02 | 上海第二工业大学 | Printed circuit board welding spot defect detection method based on digital image processing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
CN111709909B (en) | General printing defect detection method based on deep learning and model thereof | |
CN114240939B (en) | Method, system, equipment and medium for detecting appearance defects of mainboard components | |
CN114187247A (en) | Ampoule bottle printing character defect detection method based on image registration | |
CN109190625B (en) | Large-angle perspective deformation container number identification method | |
Xia et al. | SSIM-NET: Real-time PCB defect detection based on SSIM and MobileNet-V3 | |
CN109886978B (en) | End-to-end alarm information identification method based on deep learning | |
CN111445459A (en) | Image defect detection method and system based on depth twin network | |
CN111126446B (en) | Method for amplifying defect image data of robot vision industrial product | |
CN111738367B (en) | Part classification method based on image recognition | |
CN110738216A (en) | Medicine identification method based on improved SURF algorithm | |
CN116152261B (en) | Visual inspection system for quality of printed product | |
CN111415339A (en) | Image defect detection method for complex texture industrial product | |
CN111340041B (en) | License plate recognition method and device based on deep learning | |
CN111563896A (en) | Image processing method for catenary anomaly detection | |
CN111882547A (en) | PCB missing part detection method based on neural network | |
CN113780484B (en) | Industrial product defect detection method and device | |
CN113487538B (en) | Multi-target segmentation defect detection method and device and computer storage medium thereof | |
CN112926563B (en) | Fault diagnosis system for steel coil spray printing mark | |
CN113516619B (en) | Product surface flaw identification method based on image processing technology | |
CN111738310B (en) | Material classification method, device, electronic equipment and storage medium | |
CN112396582A (en) | Mask RCNN-based equalizing ring skew detection method | |
CN110969163A (en) | Method for detecting text information in image | |
CN113516193B (en) | Image processing-based red date defect identification and classification method and device | |
CN112037193A (en) | Power line feature marking method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |