CN109636824B - Multi-target counting method based on image recognition technology - Google Patents
Multi-target counting method based on image recognition technology Download PDFInfo
- Publication number
- CN109636824B CN109636824B CN201811567432.5A CN201811567432A CN109636824B CN 109636824 B CN109636824 B CN 109636824B CN 201811567432 A CN201811567432 A CN 201811567432A CN 109636824 B CN109636824 B CN 109636824B
- Authority
- CN
- China
- Prior art keywords
- image
- adhesion
- target
- area
- target object
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000005516 engineering process Methods 0.000 title claims abstract description 17
- 238000003708 edge detection Methods 0.000 claims abstract description 18
- 230000000877 morphologic effect Effects 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000003064 k means clustering Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 9
- 238000005530 etching Methods 0.000 claims description 4
- 238000005192 partition Methods 0.000 claims description 4
- 230000001629 suppression Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 238000010187 selection method Methods 0.000 claims description 3
- 239000000853 adhesive Substances 0.000 claims description 2
- 230000001070 adhesive effect Effects 0.000 claims description 2
- 230000010339 dilation Effects 0.000 claims description 2
- 230000003628 erosive effect Effects 0.000 claims description 2
- 230000007797 corrosion Effects 0.000 abstract description 6
- 238000005260 corrosion Methods 0.000 abstract description 6
- 238000012216 screening Methods 0.000 abstract description 2
- 238000004458 analytical method Methods 0.000 description 7
- 239000003990 capacitor Substances 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 229910052715 tantalum Inorganic materials 0.000 description 3
- GUVRBAGPIYLISA-UHFFFAOYSA-N tantalum atom Chemical compound [Ta] GUVRBAGPIYLISA-UHFFFAOYSA-N 0.000 description 3
- 235000002566 Capsicum Nutrition 0.000 description 2
- 239000006002 Pepper Substances 0.000 description 2
- 241000722363 Piper Species 0.000 description 2
- 235000016761 Piper aduncum Nutrition 0.000 description 2
- 235000017804 Piper guineense Nutrition 0.000 description 2
- 235000008184 Piper nigrum Nutrition 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000007621 cluster analysis Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000009826 distribution Methods 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
- 150000003839 salts Chemical class 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012549 training 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
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/30242—Counting objects in image
Abstract
The invention relates to a multi-target counting method based on an image recognition technology, which comprises the following steps: acquiring an original image by recording a target image; preprocessing an image; independent and adhesion current object screening, morphological operation is carried out on a binary image, an edge detection algorithm is adopted to identify continuous image edges, then the image is divided according to area parameters of a target object, the edge smaller than a given area threshold value can be regarded as an independent target object, the edge larger than the given threshold value can be regarded as a possible adhesion target object, and therefore distinguishing of adhesion-free and adhesion-containing areas, counting of adhesion-free areas and processing of strong adhesion areas are achieved. The multi-target counting method based on the image recognition technology and the adhesion region dividing method are carried out in the mother image by adopting filling operation after morphological corrosion operation, so that the uniqueness of an area threshold value is ensured, and the subsequent operation flow is greatly simplified.
Description
Technical Field
The invention belongs to the technical field of image recognition and machine vision, and particularly relates to a multi-target counting method based on an image recognition technology.
Background
With the development of computer technology, image recognition technology has been used in a large number of applications in production and life; the first type is various image identification (fixed target to be identified and limited types) in a general scene, and mainly identifies characters, license plates and fingerprints; another class is recognition in special scenarios, such as: in the fields of industrial automation, product sorting on a production line, medical image screening and the like (the types of targets to be identified are not fixed). The former principle is realized: for the first kind of general application scenes, the mainstream method at present mainly adopts a neural network algorithm to classify and compare samples, and can realize efficient recognition of limited target sets such as characters, license plates, fingerprints and the like through extraction and training of the characteristics of a large number of samples; for the target identification under the second type of proprietary scenes, because various targets need to be identified, the targets can be identified in a targeted manner only according to actual application scenes, and some successful cases exist; the invention mainly aims at the second kind of application condition, a series of operations are carried out on an image or video collected on site, and the operations mainly comprise: preprocessing, edge extraction, ROI division and filling, target classification, cluster analysis and statistical counting, and finally, automatic classification and counting of multi-target objects are realized.
Aiming at the second kind of application scenes, most of the second kind of application scenes are focused on the extraction and identification or defect detection of the characteristics of a single target object, the counting of medical cells and the like, firstly, a large number of samples need to be trained, secondly, the overfitting condition is easy to occur, and the images in the sample set range can be accurately counted and identified, but for unknown images outside the sample set, the identification error rate is high, and the requirements of practical application cannot be met; on the other hand, for multiple targets, large batches and adhered targets, accurate identification and counting are difficult to perform.
Disclosure of Invention
The invention aims to solve the problems and provide a multi-target counting method based on an image recognition technology, which has a simple structure and reasonable design.
The invention realizes the purpose through the following technical scheme:
the invention provides a multi-target counting method based on an image recognition technology, which comprises the following steps:
image preprocessing: collecting an original image, and carrying out gray processing on the original image to obtain a gray image; automatically adjusting a gray threshold value by adopting histogram statistics, and binarizing a gray image to obtain a binarized image;
distinguishing non-adhesion areas and adhesion areas of the target object: performing morphological operation on the binary image, and identifying continuous image edges by adopting an edge detection algorithm; setting an area threshold value according to the area parameters of the target object, dividing the image according to the area threshold value, wherein the image is regarded as an independent target object when the image is smaller than the area threshold value, and regarded as a possible target object with adhesion when the image is larger than the area threshold value, so that the non-adhesion and adhesion areas are distinguished;
counting independent target objects in the non-adhesion area;
and (3) treating the adhesion area adhesion target object: after the non-adhesion area is marked and counted, filling the original binary image to obtain a binary image only containing the target object in the adhesion area and a filling gap; then, identifying the edge of the image by using an edge detection algorithm, and calculating the area of the connected region to obtain an area sequence of the adhesion target object; the area sequences are classified by adopting a k-means clustering algorithm, and the number of the target objects is only possible to be an integer, so that the optimal selection can be carried out by judging whether the clustering centers (representing the area of one target object and two target objects …) meet the arithmetic sequence, the defect that the traditional k-means algorithm needs to appoint the number of the clustering centers in advance is effectively avoided, and the automatic counting and counting of the number of the target objects with the adhered areas are realized.
Further preferably:
the original image is acquired by adopting a color camera or a monochrome camera to acquire a target image.
The morphological operation comprises one or more erosion and/or dilation operations.
In the etching operation of the morphological operation, an appropriate operator is selected according to the size of an actual target object.
The edge detection algorithm is to carry out edge detection by using a canny operator, a sobel operator or a Laplacian operator. Further preferably, an improved Canny operator edge detection method is adopted, and the method comprises the following steps:
step 1-1, filtering the processed binary image by using mean filtering;
step 1-2, extracting edges by using first-order differential of a sobel operator template, and calculating gradient and gradient amplitude;
step 1-3, performing non-maximum suppression according to the calculated gradient and gradient amplitude;
step 1-4, adaptively selecting a highest threshold value and a lowest threshold value, and thresholding the image;
and 1-5, performing edge connection after processing according to the high and low thresholds to obtain the boundaries of all closed regions in the image and four corner points of the minimum circumscribed rectangle.
The method comprises the following steps of classifying area sequences by adopting an improved k-means clustering algorithm, carrying out optimization selection by judging whether a clustering center meets an arithmetic progression or not, and realizing automatic counting and counting of the number of target objects with adhesive areas, wherein the method comprises the following steps:
step 2-1, selecting an initial clustering central point, comprising:
(1) input area sequence set { x i At { x } i Find the maximum value x in max And the minimum value x min ；
(2) To reduce the search range of k values, the maximum number of clusters possible is determined
In the formula, A it Ceil represents rounding up for the minimum area threshold;
(3) determining k max Candidate set of initial cluster centers (k) max ≥2)；
The initial clustering centers are equally spaced, and the clusters of different area sequences meet the relation of an arithmetic progression;
step 2-2, considering the random number of the device adhesion, from this k max K values (starting from k = 2) are selected from the candidate cluster centers as theoretical cluster centers, all possible solutions are traversed, and the total is
Seed growing; for each selection method, operating a standard K-MEANS algorithm and calculating the value of Eval (K);
wherein k is the number of clusters, C i As the actual cluster center, I i As a theoretical clustering center, d i ＝|C i -I i I is the distance between the actual and theoretical cluster centers,
d when the number of clusters is k i Average value of (a); it can be known that Eval (k) is actually the variance of the clustering distance, which reflects the concentration degree of the sample to the theoretical clustering center, and the smaller the value of Eval (k), the better the concentration degree.
Step 2-3, increasing the number k of clustering partitions by 1, and repeating the steps 2-3 until k is reached max Until the end;
step 2-4, compare k max When the value of each Eval (k) is taken as the minimum value of the Eval (k), the corresponding k is the most reasonable cluster division number;
and 2-5, counting the cluster number of each adhesion target according to the cluster division number, and then calculating the number of the objects of the adhesion targets.
The invention has the beneficial effects that:
1) The adhesion region division method is carried out in the mother image by adopting filling operation after morphological corrosion operation, so that the area threshold is ensured to be unique, and the subsequent operation flow is greatly simplified;
2) The area of the target object is used as a threshold value, and a clustering analysis method is introduced at the same time, so that the method is also suitable for other clustering analysis methods;
3) The invention introduces the judgment condition that the area of the connected domain should be approximate to an arithmetic sequence as the result of the optimized clustering analysis, thereby avoiding the problem that the classified number needs to be manually specified in advance for the k-means clustering analysis method;
4) The invention has wide application range, and can meet the following requirements in other application scenes: multiple targets with adhesion and similar target objects with uncertain adhesion number can be calculated by adopting the method disclosed by the invention.
Drawings
FIG. 1 is a block diagram of the algorithm flow of the main program of the present invention;
FIG. 2 is a block diagram of the blocking area algorithm flow of the present invention;
FIG. 3 is a schematic illustration of a gray scale image acquired in accordance with the present invention;
FIG. 4 is a schematic diagram of a binarized image according to the present invention;
FIG. 5 is a graphical representation of the results of a primary etch of the present invention;
FIG. 6 is a graphical representation of the results of the secondary etch of the present invention;
FIG. 7 shows the canny edge detection results after the secondary etching according to the present invention;
FIG. 8 is a graph illustrating the parent image fill result of the present invention;
FIG. 9 is an area distribution of the adhesion zones of the present invention;
FIG. 10 is the k-means cluster analysis result of the present invention.
Detailed Description
The present application will now be described in further detail with reference to the drawings, it should be noted that the following detailed description is given for illustrative purposes only and is not to be construed as limiting the scope of the present application, as those skilled in the art will be able to make numerous insubstantial modifications and adaptations to the present application based on the above disclosure.
Example 1
As shown in fig. 1-2, a multi-target counting method based on image recognition technology, taking an image containing a plurality of tantalum capacitors as an example, according to the processing method (main program algorithm flow) described in this patent, the specific implementation process is as follows:
firstly, an image preprocessing part, after a digital image containing a plurality of tantalum capacitors is loaded by a system, graying the image to obtain a grayscale image (as shown in fig. 3), performing histogram statistics on the grayscale image, taking an average grayscale value as a threshold, a pixel value smaller than the threshold being 0, and a pixel value greater than the threshold being 255 to obtain a binarized image with a size consistent with that of an original image, and using the binarized image as a parent image of the image processing operation (a target object is white, a background is black, and reverse setting can be performed in the binarized process) (as shown in fig. 4);
secondly, identifying the outline of a single target object, firstly performing morphological corrosion operation on the image in order to count the target object, adopting a proper corrosion operator according to the size of the actual target object, wherein the operator is between 20 and 50 in the example, the corrosion operator can also take a smaller value, and the corrosion operation is executed for multiple times, and the effect is the same, the example demonstrates that the operator has the size of 20, and the result of secondary operation is executed, as shown in fig. 5 and 6, wherein the regions with non-serious adhesion are completely separated. And comparing the area (in pixel) of the connected domain with the area threshold of the target object, identifying the area smaller than the area threshold as a single target object, and counting.
For calculating the area of the connected domain, a conventional edge detection algorithm, such as Canny operator, sobel operator or Laplacian operator, may be used for edge detection. Because the image is binarized and the morphological etching operation is carried out, the identification of the target which is not adhered or independent per se is relatively easy.
The basic idea of the traditional Canny operator is to perform gaussian filtering, wherein the gaussian filtering is performed by image convolution, then an edge is extracted by a first-order differential operator, and then non-maximum suppression is performed by a gradient. Only horizontal, vertical passes are used in the first order differential operator without taking into account the 45 and 135 directions to extract edges. The high and low thresholds need to be set by themselves or the solution of double thresholds is carried out by using the sources method. And performing edge connection according to the images after the high and low threshold processing. The problems brought about are: gaussian filtering can well blur images, edge preservation is not good, the edge blurring effect is obvious, details are seriously lost, and the image is sensitive to salt and pepper noise.
The invention adopts an improved Canny operator to carry out edge detection, and comprises the following specific steps:
step 1-1, filtering the processed binary image by using mean filtering;
step 1-2, extracting edges by using first-order differential of a sobel operator template, and calculating gradient and gradient amplitude;
step 1-3, performing non-maximum suppression according to the calculated gradient and gradient amplitude;
step 1-4, adaptively selecting a highest threshold value and a lowest threshold value, and thresholding the image;
and 1-5, performing edge connection after processing according to the high and low thresholds to obtain the boundaries of all closed regions in the image and four corner points of the minimum circumscribed rectangle.
The improved detection method can overcome the technical problems that the traditional Canny operator has obvious boundary fuzzy effect, serious detail loss, sensitivity to salt and pepper noise and the like, and the result shown in figure 7 is obtained after the edge detection is carried out by adopting the improved Canny operator.
Thirdly, identifying the target in the adhesion area, and filling the identified single target (only the target in the adhesion area is left), wherein the filling operation is performed in the mother image, so that the uniqueness of the area threshold is ensured, and the filled result as shown in fig. 8 is obtained, wherein in the image, all the single targets are identified, only the adhered target objects and the filled gaps are left; entering an adhesion region processing flow, performing edge detection on the image according to the main flow, and calculating the area (taking pixels as units) of a connected domain to obtain an area sequence with an adhesion region;
in this example, taking the filled image as an example, the following connected domain areas (32 connected domains in total) are obtained, and the sequence of the connected domain areas of the blocking region is:
14810、5954、15770、15585、6504、14268、14837、28485、14856、14487、12695、13760、14154、14466、12202、13530、23277、14586、12902、13705、13539、13617、12641、13682、13989、12707、12861、26164、20286、11670、11737、19274；
the area distribution is shown in fig. 9, and k-means clustering analysis is applied to the area sequence.
The classic K-MEANS clustering algorithm (K-MEANS algorithm) can be used for analysis, but the classic K-MEANS clustering algorithm has two problems: firstly, the initial clustering center is randomly selected, so that clustering division is unstable, and a local optimal solution is easy to occur; secondly, the K-means clustering algorithm needs to manually specify the clustering division number K in advance, which greatly limits the application range of the algorithm. Therefore, the invention adopts an improved k-means clustering algorithm and aims to solve two problems: firstly, in order to ensure the consistency and the rationality of clustering division, the distance of an initial clustering center point is required to be as large as possible; secondly, the optimization evaluation of the clustering partition number K is required to be solved. The accuracy of the counting result is directly related to the two points.
The improved k-means clustering algorithm comprises the following steps:
step 2-1, selecting an initial clustering center point, comprising:
(1) input area sequence set { x i At { x } i Find the maximum value x in max And the minimum value x min ；
(2) To reduce the search range of k values, the maximum number of clusters possible is determined
In the formula, A it Ceil represents rounding up for the minimum area threshold;
in order to complete the segmentation and counting of the images, a minimum area threshold needs to be set. Because the focal length of the lens of the CCD camera and the distance between the lens and the target are fixed, the size of the imaged target object is also fixed. Therefore, different systems can properly select a proper area threshold, and the area threshold adopted by the patent is A it ∈(6200,7000)。
(3) Determining k max Candidate set of initial cluster centers (k) max ≥2)；
The initial clustering centers are equally spaced, and the clusters of different area sequences meet the relation of an arithmetic progression;
step 2-2, considering the random number of the device adhesion, from this k max Selecting k values (starting from k = 2) from the candidate cluster centers, using the selected values as theoretical cluster centers, traversing all possible access methods, and sharing the values
Seed growing; for each selection method, operating a standard K-MEANS algorithm and calculating the value of Eval (K);
wherein k is the number of clusters, C i As the actual cluster center, I i As a theoretical clustering center, d i ＝|C i -I i I is the distance between the actual and theoretical cluster centers,
d when the number of clusters is k i Average value of (a); it can be known that Eval (k) is actually the variance of the clustering distance, which reflects the concentration degree of the sample to the theoretical clustering center, and the smaller the value of Eval (k), the better the concentration degree.
Step 2-3, increasing the number k of clustering partitions by 1, and repeating the steps 2-3 until k is reached max Until the end;
step 2-4, compare k max When the value of each Eval (k) is taken as the minimum value of the Eval (k), the corresponding k is the most reasonable cluster division number;
the results are shown in FIG. 10, the area sequences are automatically classified into four categories, and the four cluster center values and the number of targets per adhesion region are shown in the table below.
Clustering center:
TABLE 1 clustering analysis results of the area sequences of the strong adhesion regions
And 2-5, calculating the number of the objects of the adhesion targets according to the clustering division number statistical result.
Therefore, counting and counting of the target objects containing a plurality of tantalum capacitors and having adhered images are completed.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.
Claims (7)
1. A multi-target counting method based on an image recognition technology is characterized by comprising the following steps:
image preprocessing: collecting an original image, and carrying out gray processing on the original image to obtain a gray image; automatically adjusting a gray threshold value by adopting histogram statistics, and binarizing a gray image to obtain a binarized image;
distinguishing non-adhesion areas and adhesion areas of the target object: performing morphological operation on the binary image, and identifying continuous image edges by adopting an edge detection algorithm; setting an area threshold value according to the area parameters of the target object, dividing the image according to the area threshold value, wherein the target object is considered to be an independent target object when the area threshold value is smaller than the area threshold value, and the target object is considered to be a possible target object with adhesion when the area threshold value is larger than the area threshold value, so that the non-adhesion and adhesion areas are distinguished;
counting independent target objects in the non-adhesion area;
and (3) treating the adhesion area adhesion target object: after the adhesion-free area is marked and counted, filling the original binary image to obtain a binary image only containing the target object in the adhesion area and a filling gap; then, identifying the edge of the image by using an edge detection algorithm, and calculating the area of the connected region to obtain an area sequence of the adhesion target object; and classifying the area sequences by adopting a k-means clustering algorithm, and performing optimal selection by judging whether a clustering center meets an arithmetic sequence or not to realize automatic counting and counting of the number of the target objects in the adhered areas.
2. The multi-target counting method based on the image recognition technology as claimed in claim 1, wherein: the original image is acquired by adopting a color camera or a monochrome camera to acquire a target image.
3. The multi-target counting method based on the image recognition technology as claimed in claim 1, wherein: the morphological operation comprises one or more erosion and/or dilation operations.
4. The multi-target counting method based on the image recognition technology as claimed in claim 3, wherein: in the etching operation of the morphological operation, an appropriate operator is selected according to the size of an actual target object.
5. The multi-target counting method based on the image recognition technology as claimed in claim 1, wherein the edge detection algorithm is edge detection using canny operator, sobel operator or Laplacian operator.
6. The multi-target counting method based on the image recognition technology according to claim 5, wherein the edge detection is a modified Canny operator edge detection method, and the method comprises the following steps:
step 1-1, filtering the processed binary image by using mean filtering;
step 1-2, extracting edges by using first-order differential of a sobel operator template, and calculating gradient and gradient amplitude;
step 1-3, performing non-maximum suppression according to the calculated gradient and gradient amplitude;
step 1-4, adaptively selecting a highest threshold value and a lowest threshold value, and thresholding the image;
and 1-5, performing edge connection after processing according to the high and low thresholds to obtain the boundaries of all closed regions in the image and four corner points of the minimum circumscribed rectangle.
7. The multi-target counting method based on the image recognition technology as claimed in claim 1, wherein an improved k-means clustering algorithm is adopted to classify area sequences, and optimization selection is performed by judging whether a clustering center meets an arithmetic progression or not, so that automatic counting and counting of the number of target objects with adhesive areas are realized, and the method comprises the following steps:
step 2-1, selecting an initial clustering center point, comprising:
(2) input area sequence set { x i At { x } i Find the maximum value x in max And the minimum value x min ；
(2) To reduce the search range of k values, the maximum number of clusters possible is determined
In the formula, A it Ceil represents rounding up for the minimum area threshold;
(3) determining k max Candidate set of initial cluster centers (k) max ≥2)；
The initial clustering centers are equally spaced, and the clusters of different area sequences meet the relation of an arithmetic progression;
step 2-2 from this k max K values (starting from k = 2) are selected from the candidate cluster centers as theoretical cluster centers, all possible solutions are traversed, and the total is
Seed growing; for each selection method, operating a standard K-MEANS algorithm and calculating the value of Eval (K);
wherein k is the number of clusters, C i As the actual cluster center, I i As a theoretical clustering center, d i ＝|C i -I i I is the distance between the actual and theoretical cluster centers,
d is when the number of clusters is k i Average value of (a);
step 2-3, increasing the number k of clustering partitions by 1, and repeating the steps 2-3 until k is reached max Until the end;
step 2-4, compare k max When the value of each Eval (k) is taken as the minimum value of the Eval (k), the corresponding k is the most reasonable cluster division number;
and 2-5, counting the cluster number of each adhesion target according to the cluster division number, and then calculating the number of the objects of the adhesion targets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811567432.5A CN109636824B (en) | 2018-12-20 | 2018-12-20 | Multi-target counting method based on image recognition technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811567432.5A CN109636824B (en) | 2018-12-20 | 2018-12-20 | Multi-target counting method based on image recognition technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109636824A CN109636824A (en) | 2019-04-16 |
| CN109636824B true CN109636824B (en) | 2022-10-11 |
Family
ID=66076126
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811567432.5A Active CN109636824B (en) | 2018-12-20 | 2018-12-20 | Multi-target counting method based on image recognition technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109636824B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264463A (en) * | 2019-06-25 | 2019-09-20 | 北京实验工厂有限责任公司 | A kind of material counting method based on matlab image procossing |
| CN112307828A (en) * | 2019-07-31 | 2021-02-02 | 梅特勒-托利多(常州)测量技术有限公司 | Count verification device, count system and method |
| CN112819745B (en) * | 2019-10-31 | 2023-02-28 | 合肥美亚光电技术股份有限公司 | Nut kernel center worm-eating defect detection method and device |
| CN111058182A (en) * | 2019-12-25 | 2020-04-24 | 杭州晶一智能科技有限公司 | Yarn state detection method based on projection area statistics |
| CN111583255B (en) * | 2020-05-19 | 2023-06-13 | 东莞市瑞图新智科技有限公司 | Multi-domain-oriented high-adaptability identification and counting method |
| CN111753851B (en) * | 2020-07-01 | 2022-06-07 | 中国铁路设计集团有限公司 | Railway snow depth and wind and snow migration track monitoring method and system based on image processing |
| CN113313692B (en) * | 2021-06-03 | 2023-04-25 | 广西大学 | Automatic banana young plant identification and counting method based on aerial visible light image |
| CN115345879B (en) * | 2022-10-18 | 2023-03-24 | 济宁康盛彩虹生物科技有限公司 | Method for analyzing corrosion degree of inner wall of autoclave and predicting service life of autoclave based on image |
| CN115546462A (en) * | 2022-12-01 | 2022-12-30 | 南京维拓科技股份有限公司 | Method for extracting shape features of product and counting based on image recognition |
| CN115619789B (en) * | 2022-12-20 | 2023-03-31 | 山西蒲谷香农业科技有限公司 | Pill forming control method for production of black sesame pills |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1282076A2 (en) * | 2001-07-31 | 2003-02-05 | Canon Kabushiki Kaisha | Self-adaptive binarization of digital images |
| CN102043950A (en) * | 2010-12-30 | 2011-05-04 | 南京信息工程大学 | Vehicle outline recognition method based on canny operator and marginal point statistic |
| WO2014067296A1 (en) * | 2012-11-05 | 2014-05-08 | 深圳市恩普电子技术有限公司 | Method for identifying, tracing and measuring internal and external membranes of vessel |
| CN107895376A (en) * | 2017-12-11 | 2018-04-10 | 福州大学 | Based on the solar panel recognition methods for improving Canny operators and contour area threshold value |
-
2018
- 2018-12-20 CN CN201811567432.5A patent/CN109636824B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1282076A2 (en) * | 2001-07-31 | 2003-02-05 | Canon Kabushiki Kaisha | Self-adaptive binarization of digital images |
| CN102043950A (en) * | 2010-12-30 | 2011-05-04 | 南京信息工程大学 | Vehicle outline recognition method based on canny operator and marginal point statistic |
| WO2014067296A1 (en) * | 2012-11-05 | 2014-05-08 | 深圳市恩普电子技术有限公司 | Method for identifying, tracing and measuring internal and external membranes of vessel |
| CN107895376A (en) * | 2017-12-11 | 2018-04-10 | 福州大学 | Based on the solar panel recognition methods for improving Canny operators and contour area threshold value |
Non-Patent Citations (2)
| Title |
|---|
| PMMW金属目标图像处理与计数技术实现;林晓敏等;《电光与控制》;20120701(第07期);全文 * |
| 基于局部期望阈值分割的图像边缘检测算法;刘占;《计算机与现代化》;20160812(第08期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109636824A (en) | 2019-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109636824B (en) | Multi-target counting method based on image recognition technology | |
| CN110334706B (en) | Image target identification method and device | |
| CN110443128B (en) | Finger vein identification method based on SURF feature point accurate matching | |
| CN116205919B (en) | Hardware part production quality detection method and system based on artificial intelligence | |
| Tonti et al. | An automated approach to the segmentation of HEp-2 cells for the indirect immunofluorescence ANA test | |
| WO2021109697A1 (en) | Character segmentation method and apparatus, and computer-readable storage medium | |
| CN111027546B (en) | Character segmentation method, device and computer readable storage medium | |
| CN101122953A (en) | Picture words segmentation method | |
| CN111145209A (en) | Medical image segmentation method, device, equipment and storage medium | |
| CN114399522A (en) | High-low threshold-based Canny operator edge detection method | |
| CN111476804B (en) | Efficient carrier roller image segmentation method, device, equipment and storage medium | |
| Pandit et al. | Literature review on object counting using image processing techniques | |
| Öztürk et al. | Comparison of HOG, MSER, SIFT, FAST, LBP and CANNY features for cell detection in histopathological images | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN107704864B (en) | Salient object detection method based on image object semantic detection | |
| Sujatha et al. | Enhancement of segmentation and feature fusion for apple disease classification | |
| CN113450373A (en) | Optical live image-based real-time discrimination method for characteristic events in carrier rocket flight process | |
| Tong et al. | Document image binarization based on NFCM | |
| Abraham | Plasmodium detection using simple CNN and clustered GLCM features | |
| CN110276260B (en) | Commodity detection method based on depth camera | |
| Umamaheswari et al. | Fuzzy-C means segmentation of lymphocytes for the identification of the differential counting of WBC | |
| CN110619273B (en) | Efficient iris recognition method and recognition device | |
| Ambata et al. | Foreground background separation and tracking | |
| CN113850792A (en) | Cell classification counting method and system based on computer vision | |
| Gim et al. | A novel framework for white blood cell segmentation based on stepwise rules and morphological features |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20231008 Address after: Room 13, Building 38, Suning Yaju, No. 258 Baota Road, Chunxi Street, Gaochun District, Nanjing City, Jiangsu Province, 210000 Patentee after: Jiangsu Dingwei Technology Co.,Ltd. Address before: 238000 Hechao Economic Development Zone, Hefei City, Anhui Province Patentee before: CHAOHU University |
|
| TR01 | Transfer of patent right |