CN109658395B - Optic disc tracking method and system and eye fundus collection device - Google Patents
Optic disc tracking method and system and eye fundus collection device Download PDFInfo
- Publication number
- CN109658395B CN109658395B CN201811485787.XA CN201811485787A CN109658395B CN 109658395 B CN109658395 B CN 109658395B CN 201811485787 A CN201811485787 A CN 201811485787A CN 109658395 B CN109658395 B CN 109658395B
- Authority
- CN
- China
- Prior art keywords
- disc
- fundus image
- frame
- judgment
- current
- 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 39
- 238000004364 calculation method Methods 0.000 claims description 33
- 238000012545 processing Methods 0.000 claims description 18
- 230000003287 optical effect Effects 0.000 claims description 12
- 238000006467 substitution reaction Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000003333 near-infrared imaging Methods 0.000 abstract description 5
- 210000003128 head Anatomy 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 210000003733 optic disk Anatomy 0.000 description 2
- 210000004204 blood vessel Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 210000002189 macula lutea Anatomy 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000003068 static effect Effects 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
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- 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
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Eye Examination Apparatus (AREA)
Abstract
The invention discloses a video disc tracking method, which comprises the steps of calculating the video disc centroid position of each frame of fundus image in a near-infrared fundus video stream; extracting initial former N frames of fundus images and the centroid position of the optic disc, and determining the initial frame of the current judgment queue; calculating the variance between the disc centroid position of the current frame fundus image and the disc centroid position of the previous effective frame fundus image, and the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc; judging whether the fundus image of the current frame is a noise point according to the variance and the reference distance; when noise is determined, the current frame fundus image is removed; when judging that the current frame is not a noise point, caching the current frame fundus image; and continuously judging whether the fundus image of the next frame is noisy or not until the length of the current judgment queue reaches a threshold value, eliminating the earliest frame according to a first-in first-out principle and introducing a new frame. The method solves the problem that the traditional optic disc positioning method is not suitable for dynamic and near-infrared imaging fundus images.
Description
Technical Field
The invention relates to the technical field of medical image processing, in particular to a video disc tracking method and system and an eye fundus acquisition device.
Background
The optic disc is one of the key elements of the fundus imaging graph, and is used as a key element which is depended on by fundus lesion medical diagnosis together with blood vessels, macula lutea and retina. Most of the currently used methods and devices for performing image recognition on the optic disc analyze a static fundus image after visible light imaging, extract the optic disc, and assist medical diagnosis. But not for dynamic, near-infrared imaging fundus images.
Disclosure of Invention
Based on this, it is necessary to provide a disc tracking method and system and an ocular fundus collection device to solve the problem that the conventional disc identification and positioning method and device are not suitable for dynamic near-infrared imaging fundus images.
In view of the above, the present invention provides a method for tracking an optical disc, comprising the steps of:
acquiring a collected near-infrared fundus video stream, and calculating the position of the centroid of a optic disc of each frame of fundus image in the near-infrared fundus video stream;
extracting initial first N frames of fundus images and the centroid position of the optic disc in the near-infrared video stream, and determining an initial frame of a current judgment queue from the first N frames of fundus images according to the extracted centroid position of the optic disc; wherein the fixed length threshold of the current decision queue is n;
calculating the variance of the disc centroid position of the current frame fundus image and the disc centroid position of the previous effective frame fundus image and the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc for the current frame fundus image newly captured in the current judgment queue; wherein, the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc is a reference distance;
judging whether the optic disc centroid position of the current frame fundus image is a noise point or not according to the calculated variance and the reference distance; and when the optic disc center of mass of the current frame eye fundus image is judged to be a noise point, the current frame eye fundus image is removed; when the disc centroid position of the current frame fundus image is judged not to be a noise point, caching the current frame fundus image and the disc centroid position thereof;
continuously capturing the next frame of fundus images in the current judgment queue and judging until the length of the current judgment queue reaches a threshold value n, and eliminating the frame at the head of the queue according to a first-in first-out principle while entering each frame of fundus images captured subsequently into the current judgment queue.
In one embodiment, the extracting the initial previous N frames of fundus images and the disc centroid position in the near-infrared video stream, and determining the initial frame of the current decision queue from the previous N frames of fundus images according to the extracted disc centroid position, comprises the steps of:
extracting the first N frames of fundus images and the corresponding N positions of the centroid of the optic disc from the near-infrared fundus video stream;
respectively carrying out average value calculation on the abscissa and the ordinate of the N optic disc mass center positions to obtain an average mass center position;
calculating the distance between each of the mass center positions of the optic disc and the average mass center position to obtain N average distances;
sorting the N average distances in size, and removing a preset number of average distances in a descending order to obtain N-Q reasonable average distances; wherein Q is a preset number;
and selecting the maximum average distance from the reasonable average distances as a boundary value, comparing the maximum average distance with the boundary value as a reference according to the time sequence reverse order of the previous N frames, and acquiring a frame of which the distance between the first video disc centroid position and the average centroid position is less than or equal to the boundary value as the initial frame.
In one embodiment, the value range of the preset number Q is: q is more than or equal to N10% and less than or equal to N30%.
In one embodiment, the determining whether the disc centroid position of the current fundus image is noisy based on the calculated variance and the reference distance includes:
substituting the variance and the reference distance into a decision formula: p (m) a 1 ×(d m -d Reference(s) )+a 2 ×l m Calculating to obtain a corresponding judgment value; wherein P (m) is a calculated judgment value, a 1 Is a variance influence coefficient, a 2 As a reference distance influence coefficient, d m Is the variance, d Datum Is a reference variance,/ m Is the reference distance;
judging whether the judgment value is larger than or equal to a preset value;
when the judgment value is judged to be larger than or equal to the preset value, judging that the optic disc centroid position of the current fundus image is a noise point; and when the judgment value is judged to be smaller than the preset value, judging that the position of the centroid of the optic disc of the current fundus image is not a noise point.
In one embodiment, the method further comprises the following steps in the process of continuously capturing the fundus image of the next frame and judging:
counting the number of noise points judged currently;
and when counting that the centroid positions of the optic discs corresponding to the current continuous b frames of fundus images are all noise points, cancelling the judgment of the next frame of fundus image and resetting the judgment queue.
Correspondingly, based on the same invention concept, the invention also provides a video disc tracking system, which comprises a video disc mass center position calculating module, a judging queue determining module, a judging parameter calculating module and a noise point judging and processing module;
the optical disk centroid position calculation module is used for acquiring the acquired near-infrared fundus video stream and calculating the optical disk centroid position of each frame of fundus image in the near-infrared fundus video stream;
the judgment queue determining module is used for extracting initial first N frames of fundus images and the centroid position of the optic disc in the near-infrared video stream, and determining an initial frame of a current judgment queue from the first N frames of fundus images according to the extracted centroid position of the optic disc; wherein the fixed length threshold of the current decision queue is n;
the judgment parameter calculation module is used for calculating the variance between the disc centroid position of the current frame fundus image and the disc centroid position of the previous effective frame fundus image and the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc aiming at the current frame fundus image newly captured in the current judgment queue; the distance between the optic disc mass center position of the current frame fundus image and the optic disc mass center position of the reference optic disc is a reference distance;
the noise point judgment processing module is used for judging whether the optic disc centroid position of the current frame fundus image is a noise point according to the variance and the reference distance obtained by calculation; and when the optic disc center of mass of the current frame eye fundus image is judged to be a noise point, the current frame eye fundus image is removed; when the disc centroid position of the current frame of the fundus image is judged not to be a noise point, caching the current frame of the fundus image and the disc centroid position thereof to an output queue;
the noise point judgment processing module is further configured to skip to the noise point judgment module after the current frame fundus image is removed or cached, continue to capture and judge the next frame fundus image in the current judgment queue by the noise point judgment module until the length of the current judgment queue reaches the fixed threshold n, enter the current judgment queue for each frame fundus image captured subsequently, and simultaneously eliminate the frame at the head of the current judgment queue according to a first-in first-out principle.
In one embodiment, the decision queue determining module comprises an extracting submodule, an average value calculating submodule, an average distance calculating submodule, a sorting submodule and a selecting and comparing submodule;
the extraction submodule is used for extracting all fundus images and the corresponding optic disc centroid positions from the near-infrared fundus video stream;
the average value calculation submodule is used for respectively carrying out average value calculation on the abscissa and the ordinate of the mass center positions of the first N optical disks to obtain an average mass center position;
the average distance calculation submodule is used for calculating the distance between each of the mass center positions of the optic disc and the average mass center position to obtain N average distances;
the sorting submodule is used for sorting the N average distances in size and eliminating the maximum average distances in a preset number from large to small to obtain N-Q reasonable average distances; wherein Q is a preset number;
and the selection and comparison submodule is used for selecting the maximum average distance from the reasonable average distances as a boundary value, comparing the maximum average distance with the boundary value as a reference according to the time sequence reverse order of the previous N frames, and acquiring a frame of which the distance between the first video disc centroid position and the average centroid position is smaller than or equal to the boundary value as the initial frame.
In one embodiment, the noise judgment processing module comprises a numerical value substitution submodule and a judgment value judgment submodule;
wherein the numerical value substitution submodule is configured to substitute the variance and the reference distance into a decision formula: p (m) ═ a 1 ×(d m -d Datum )+a 2 ×l m Calculating to obtain a corresponding judgment value; wherein P (m) is a calculated judgment value, a 1 Is a variance influence coefficient, a 2 As a reference distance influence coefficient, d m Is the variance, d Datum Is a reference variance,/ m Is the reference distance;
the judgment value judgment submodule is used for judging whether the judgment value is greater than or equal to a preset value;
the judgment value judgment submodule is also used for judging that the position of the optic disc centroid of the current fundus image is a noise point when the judgment value is judged to be larger than or equal to the preset value; and when the judgment value is judged to be smaller than the preset value, judging that the optic disc centroid position of the current fundus image is not a noise point.
In one embodiment, the method further comprises a noise number statistic module;
the noise number counting module is used for counting the number of currently determined noise in the judgment process that whether the optic disc centroid position of the next frame of fundus image in the current judgment queue is a noise or not by the noise judgment processing module;
the noise number counting module is further used for canceling the judgment of the next frame of fundus image when the disc mass center positions corresponding to the current continuous b frame of fundus image are counted to be noise, skipping to the judgment queue determining module, and resetting the judgment queue by the judgment queue determining module.
Accordingly, based on the same inventive concept, the invention also provides a fundus collection device comprising the optic disc tracking system as described in any one of the above.
The optic disc tracking method comprises the steps of calculating the optic disc mass center position of each frame of eye fundus image in a near-infrared eye fundus video stream by adopting a near-infrared eye fundus optic disc positioning algorithm, judging whether each frame of eye fundus image and the optic disc mass center position thereof in a determined judging queue are noise points or not, wherein when judging whether the frame of eye fundus image and the optic disc mass center position thereof are the noise points or not, calculating the variance between the optic disc mass center position of the current eye fundus image and the optic disc mass center position of the eye fundus image of the previous effective frame and the reference distance between the optic disc mass center position of the current eye fundus image and the optic disc mass center position of the reference optic disc, judging according to the calculated variance and the reference distance, effectively eliminating the interference of the noise points in an output sequence, realizing the tracking and positioning of the optic disc position in the eye fundus image of dynamic and near-infrared imaging, and finally effectively solving the problem that the traditional optic disc identification and positioning method and device are not suitable for the dynamic eye fundus image, Problem of fundus map for near infrared imaging.
Drawings
FIG. 1 is a flow chart of an embodiment of a disc tracking method of the present invention;
FIG. 2 is a process diagram of another embodiment of the disc tracking method of the present invention;
FIG. 3 is a schematic block diagram of an embodiment of a disc tracking system according to the present invention;
FIG. 4 is a schematic diagram of another embodiment of the disc tracking system of the present invention.
Detailed Description
In order to make the technical solution of the present invention clearer, the present invention is further described in detail with reference to specific embodiments below. It should be noted, however, that the following description includes various specific details to assist in understanding, but these details are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the various embodiments described herein can be made without departing from the scope and spirit of the present disclosure. In addition, descriptions of well-known functions and constructions may be omitted for clarity and conciseness.
The terms and words used in the following description and claims are not limited to literature meanings, but are used only by the inventor to enable the disclosure to be clearly and consistently understood. Accordingly, it should be apparent to those skilled in the art that the following descriptions of the various embodiments of the present disclosure are provided for illustration only and not for the purpose of limiting the disclosure as defined by the appended claims and their equivalents.
It is to be understood that the singular forms also include the plural reference unless the context clearly dictates otherwise. Thus, for example, reference to a "component surface" includes reference to one or more such surfaces.
Referring to fig. 1, as an embodiment of the optic disc tracking method of the present invention, it first includes step S100, acquiring a captured near-infrared fundus video stream, and calculating the disc centroid position of each frame of fundus image in the near-infrared fundus video stream. Here, it should be noted that, since the acquired near-infrared fundus video stream includes several frames of fundus images, after the acquisition of the near-infrared fundus video stream, it is necessary to perform disc-location tracking on each frame of fundus image in the video stream. In the localization tracking process, the disc centroid position of each frame of fundus image needs to be determined first. Preferably, the disc centroid position of each frame of fundus image is calculated using a near-infrared disc positioning algorithm.
Specifically, the near-infrared optic disc positioning algorithm can be realized through the following steps. First, the g channel of rgb is extracted from the near-infrared fundus image. Then, an open source circle search algorithm is used for processing the fundus image of the g channel, and parameters such as the mass center, the perimeter, the radius and the like of a plurality of extracted possible circles are obtained. And then, comparing the values of the perimeter and the radius with a reasonable threshold value, and screening out 1 optimal solution to obtain the centroid of the optic disc.
After the disk centroid position of each frame of fundus image in the video stream is calculated by the method, step S200 can be executed to extract the initial first N frames of fundus images and the disk centroid position of each frame of fundus image in the near-infrared fundus video stream, and determine the initial frame of the current determination queue from the first N frames of fundus images according to the disk centroid position of each extracted frame of fundus image.
Specifically, when the initial first N frames of fundus images and the corresponding N disc centroid positions are extracted from the video stream, and the current determination queue is determined from the first N frames of fundus images according to the extracted disc centroid positions, first, in step S210, the initial first N frames of fundus images and the disc centroid position corresponding to each frame of fundus image are extracted from the near-infrared fundus video stream. At this time, the extracted data includes N frames of fundus images and N disc centroid positions. Wherein, the N optic disc centroid positions can be characterized in a two-dimensional coordinate mode. That is, the centroid positions of the optical disk corresponding to the extracted first N frames of fundus images are respectively: (x) 1 ，y 1 )、(x 2 ，y 2 )、(x 3 ，y 3 )、……、(x N ，y N )。
After the initial first N frames of fundus images and the corresponding centroid positions of the optical discs are extracted, step S220 may be executed to perform average calculation on the abscissa and ordinate of the extracted centroid positions of the N optical discs, respectively, to obtain the corresponding average centroid position. That is, according to the calculation formula:
calculating to obtain an average centroid position: (x) Average ，y Average out )。
The average centroid position (x) of the fundus images of the first N frames is obtained by the above calculation formula Average out ，y Average out ) Then, in step S230, the distance between each of the N optic disc centroid positions and the average centroid position is calculated, so as to obtain N average distances. That is, according to the distance calculation formula:
n average distances are calculated: d is a radical of 1 average 、d 2 average 、d 3 average of 、……、d m mean 、……、d N average . Wherein m is 1 to N, d m mean Characterizing the mth disc centroid position (x) of the N disc centroid positions m ，y m ) And the mean centroid position (x) Average out ，y Average ) The average distance between them.
And then step S240 is executed, the N calculated average distances are sorted, and N-Q reasonable average distances are obtained after the preset number of maximum average distances are sequentially removed from the maximum to the minimum. That is, for the N calculated average distances: d 1 average 、d 2 average 、d 3 average of 、……、d m mean 、……、d N average Sorting the sizes, such as: and sequentially arranging the distance data according to the sequence from large to small so as to obtain a sequence of average distance data with gradually-reduced values. Then, the average distances of a preset number Q (for example, the value of the preset number Q is 30% of N) in the average distance number series are sequentially removed from large to small, so that the remaining (N-Q) average distances which are not removed are kept as a reasonable average distance. Further, it should be noted here that the value range of the preset number Q is: q is more than or equal to N10% and less than or equal to N30%.
Then, step S250 is executed to select the maximum average distance from the reasonable average distances as a boundary value, and compare the maximum average distance with the boundary value as a reference according to the time sequence of the previous N frames in reverse order, and obtain a frame with the distance between the first video disc centroid position and the average centroid position smaller than or equal to the boundary value as the initial frame of the current determination queue.
Here, it should be noted that the fixed length threshold of the current decision queue is n. Wherein N is less than N-Q.
After the initial frame of the current judgment queue is determined through the steps, the judgment whether the centroid position of the optic disc of the newly captured fundus image of the current frame is a noise point or not can be carried out. In an embodiment of the optic disc tracking method of the present invention, referring to fig. 2, the criterion for determining whether the optic disc centroid position of the fundus image is noise is mainly implemented by calculating the variance between the optic disc centroid position of the current frame fundus image currently being determined and the optic disc centroid position of the previous effective frame fundus image, performing a rationality comparison between the calculated variance and the reference variance, and further performing a rationality evaluation according to the distance between the optic disc centroid position of the current frame fundus image and the reference optic disc centroid position.
Therefore, based on the above judgment criteria, in another embodiment of the disc tracking method of the present invention, it comprises: step S310, calculating the variance of the disc centroid position of the current fundus image and the disc centroid position of the fundus image of the previous effective frame in the current determination queue. Specifically, the formula for calculating the variance is as follows:
wherein xm and ym are respectively the abscissa and ordinate of the optic disc centroid position of the current frame fundus image, and x q ，y q Respectively the abscissa and the ordinate of the centroid position of the optic disc of the fundus image of the previous effective frame. The method can be obtained by substituting the optic disc centroid position of the current frame eye fundus image and the optic disc centroid position of the previous effective frame eye fundus image into a calculation variance formula, and is simple in calculation and convenient to realize.
Meanwhile, by step S320 as well, the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc is calculated as the reference distance. Here, it should be noted that the range of values of the reference optic disc centroid position is: and (3) taking the center point of the center of the fundus image as a 0 coordinate of the xy axis, and the diameter distance of the optic disc as dp, so that the centroid coordinate of the reference optic disc is (-dp, -1/2dp) for the left eye, and the centroid coordinate of the reference optic disc is (dp, -1/2dp) for the right eye.
Further, in step S400, it is determined whether or not the disc centroid position of the current fundus image is a noise point based on the calculated variance and the reference distance.
Specifically, when the determination as to whether or not the current disc centroid position of the fundus image is noisy is made based on the calculated variance and reference distance in step S400, first the variance and reference distance are substituted into the determination formula in step S410：P(m)＝a 1 ×(d m -d Datum )+a 2 ×l m And calculating to obtain a corresponding judgment value. Wherein P (m) is the calculated judgment value. a is 1 The variance influence coefficient has the following value range: 0.85 to 0.95, preferably 0.93. a is a 2 The reference distance influence coefficient is defined as the following value range: 0.05 to 0.15, preferably 0.07. d m Is the variance between the disc centroid position of the mth frame fundus image (i.e., the current frame fundus image) and the disc centroid position of the previous valid frame fundus image. d is a radical of Datum The value range of the preset reference variance can be set to be 0.5-1.5 optic disc diameters, and is preferably 0.8 optic disc diameters. l. the m The distance (i.e., the reference distance) between the disc centroid position of the current fundus image and the disc centroid position of the reference disc is between 2 and 4 disc diameters, preferably 2.7 disc diameters.
When the corresponding judgment value P (m) is obtained through calculation, the judgment value P (m) can be judged. Specifically, step S420 is executed to determine whether the determination value p (m) is greater than or equal to a predetermined value. When the determination value p (m) is judged to be greater than or equal to the preset value, the optic disk centroid position of the current fundus image is judged to be noise. When the judgment value P (m) is judged to be smaller than the preset value, the optic disc centroid position of the current fundus image is judged not to be a noise point. Wherein, the value of the preset value is preset, and the preset value is preferably 0.9 optic disc diameter.
Therefore, by taking the two parameters of the variance between the optic disc mass center position of the current frame eye fundus image and the optic disc mass center position of the previous effective frame eye fundus image and the distance between the optic disc mass center position of the current frame eye fundus image and the optic disc mass center position of the reference optic disc (as the reference distance) as the basis for noise judgment, the noise for optic disc positioning can be effectively screened out, and the noise is eliminated in an output sequence, so that the relative continuity and rationality of optic disc marking coordinates are ensured, and the effective tracking of the optic disc is finally realized.
Here, it is to be noted that, in the above-described step of calculating the variance, the previous valid frame fundus image of the current frame fundus image refers to a frame fundus image which is located in front of the current frame fundus image and is adjacent to the current frame fundus image and is not a noise.
Further, as another embodiment of the optic disc tracking method according to the present invention, after the determination as to whether the current fundus image is noisy is completed through the above steps, step S500 may be performed to continue capturing the fundus image of the next frame in the current determination queue and to perform the determination as to whether the captured fundus image of the next frame is noisy.
In the determination process of whether the fundus image of the next frame is noisy, it is preferable that the method further includes a step S500' of counting the number of currently determined noisy points. When the statistics shows that the centroid positions of the optic discs of the current continuous b-frame fundus images are all noise points, the fact that the current determination queue determined in the front is unreasonable is shown, and the judgment can not be carried out due to poor imaging quality caused by shooting light leakage and the like, so that the judgment on whether the fundus image of the next frame is noise points is cancelled at the moment, and the resetting of the determination queue is carried out. Wherein the value of b is preferably 5.
Furthermore, after the length of the current judgment queue reaches the threshold value n, for each newly captured frame of fundus image, according to a first-in first-out principle, one frame of fundus image with the earliest time is eliminated from the head of the current judgment queue until the disc centroid position of each frame of fundus image in the collected near-infrared fundus video stream is judged whether to be noise or not.
Therefore, the optic disc tracking method of the invention, through calculating the optic disc mass center position of each frame of the eye fundus image of the near infrared eye fundus video stream, and caching the optic disc mass center positions of continuous N frames through the queue, the determination of the initial frame of the queue is carried out on the cached continuous N frames of the optic disc mass center positions, and after the initial frame of the queue is determined, the variance and the reference distance are calculated on a plurality of continuous frames of eye fundus images captured subsequently, so that the weight calculation is carried out according to the two parameters of the variance and the reference distance obtained by calculation, the noise point screening and the elimination are carried out according to the weight, the relative continuity and the reasonability of the output optic disc mark coordinates are ensured, and the aim of optic disc tracking is realized.
Correspondingly, in order to realize any one of the video disc tracking methods, the invention also provides a video disc tracking system. Since the working principle of the video disc tracking system provided by the invention is the same as or similar to that of the video disc tracking method provided by the invention, repeated descriptions are omitted.
Referring to fig. 3, as an embodiment of the disc tracking system 100 of the present invention, it comprises a disc centroid position calculation module 110, a decision queue determination module 120, a decision parameter calculation module 130 and a noise point judgment processing module 140.
The optic disc centroid position calculating module 110 is configured to acquire the acquired near-infrared fundus video stream and calculate a optic disc centroid position of each frame of fundus image in the near-infrared fundus video stream. A decision queue determining module 120, configured to extract an initial previous N frames of fundus images in the near-infrared video stream and a disc centroid position thereof, and determine an initial frame of a current decision queue from the previous N frames of fundus images according to the extracted disc centroid position; the fixed length threshold of the current decision queue is N, and N is less than or equal to N. A decision parameter calculation module 130, configured to calculate, for a current frame fundus image newly captured in the current decision queue, a variance between the disc centroid position of the current frame fundus image and the disc centroid position of a previous effective frame fundus image, and a distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc; wherein, the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc is the reference distance. A noise point judgment processing module 140, configured to judge whether the optic disc centroid position of the current frame fundus image is a noise point according to the calculated variance and the reference distance; when the optic disk centroid position of the current frame eye fundus image is judged to be a noise point, the current frame eye fundus image is removed; and when the disc centroid position of the current frame fundus image is judged not to be a noise point, caching the current frame fundus image. The noise point judgment processing module 140 is further configured to skip to the noise point judgment module after the current frame fundus image is removed or cached, continue to capture and judge the next frame fundus image in the current judgment queue by the noise point judgment module until the length in the current judgment queue reaches the fixed threshold n, enter the current judgment queue for each frame fundus image captured subsequently, and simultaneously eliminate the frame at the head of the current judgment queue according to a first-in first-out principle.
Preferably, in the optical disc tracking system 100 of the present invention, the decision queue determining module 120 specifically includes an extracting sub-module, an average value calculating sub-module, an average distance calculating sub-module, a sorting sub-module and a selecting and comparing sub-module (not shown in the figure).
The extraction submodule is used for extracting the front N frames of fundus images and the corresponding N video disc centroid positions from the near-infrared fundus video stream.
And the average value calculation submodule is used for respectively carrying out average value calculation on the abscissa and the ordinate of the mass center positions of the front N optic discs to obtain the average mass center position.
And the average distance calculation submodule is used for calculating the distance between each of the mass center positions of the N discs and the average mass center position to obtain N average distances.
The sorting submodule is used for sorting the N average distances in size and eliminating the maximum average distances in a preset number from large to small to obtain N-Q reasonable average distances; wherein Q is a preset number.
And the selection comparison submodule is used for selecting the maximum average distance from the reasonable average distances as a boundary value, comparing the maximum average distance with the boundary value as a reference according to the time sequence reverse order of the previous N frames, and acquiring a frame as the initial frame, wherein the distance between the first video disc centroid position and the average centroid position is less than or equal to the boundary value.
Further, referring to fig. 4, as another embodiment of the disc tracking system 100 of the present invention, the noise judgment processing module 140 includes a numerical substitution sub-module 141 and a judgment value judgment sub-module 142. Wherein, the numerical value substitution submodule 141 is configured to substitute the variance and the reference distance into the decision formula: p (m) a 1 ×(d m -d Datum )+a 2 ×l m Calculating to obtain a corresponding judgment value; wherein P (m) is a calculated judgment value, a 1 Is a variance influence coefficient, a 2 As reference distance influence coefficient, d m Is the variance, d Datum Is a reference variance,/ m Is the reference distance.
And a decision value judging submodule 142 for judging whether the decision value is greater than or equal to a preset value. The decision value judging submodule 142 is further configured to, when the decision value is judged to be greater than or equal to the preset value, judge the position of the optic disc centroid of the current frame fundus image as a noise point; and when the judgment value is smaller than the preset value, judging that the position of the optic disc centroid of the current frame fundus image is not a noise point.
Further, in another embodiment of the disc tracking system 100 of the present invention, a noise count module (not shown) is further included. The noise number counting module is configured to count the number of currently determined noise in the process that the noise determination processing module 140 continues to determine whether the optic disc centroid position of the next frame of fundus image in the current determination queue is a noise.
Here, it should be noted that the noise number statistics module is further configured to cancel the determination on the fundus image of the next frame when it is counted that the centroid positions of the optical disc corresponding to the current continuous b-frame fundus images are all noise, and jump to the determination queue determination module 120, and the determination queue determination module 120 resets the determination queue.
The optic disc tracking system 100 of the invention, through setting up the optic disc centroid position calculating module 110, calculate the optic disc centroid position of each frame of eyeground image of the near-infrared eyeground video stream by the optic disc centroid position calculating module 110, and through setting up and judging the queue confirming module 120, judge the confirming module 120 to judge the confirming of the queue to the consecutive N frames of optic disc centroid positions cached, judge the module 130 to judge several frames of eyeground images in the queue to carry on the calculation of variance and reference distance to confirm at present after judging the queue, thus judge processing module 140 carry on the weight calculation according to variance and reference distance two parameters that calculate through the noise, carry on the screening and rejecting of the noise according to the weight, guarantee that the optic disc mark coordinate outputted is relatively continuous, rational, have realized the goal that the optic disc tracks.
In addition, based on any of the above-mentioned disc tracking systems 100, the present invention also provides a fundus collection apparatus including the disc tracking system 100 as described in any of the above-mentioned.
By applying any video disc tracking system to the fundus collection device, the fundus collection device can effectively process a plurality of frames of near-infrared fundus preview video streams per second, and perform video disc positioning, denoising and marking on each frame of fundus images, thereby realizing the video disc tracking effect under near-infrared light and assisting a photographer to shoot the fundus images with the video disc positions meeting the judgment standard.
In addition, it should be noted that, the technical features of the above embodiments may be arbitrarily combined, and for the sake of simplicity of description, all possible combinations of the technical features in the above embodiments are not described, however, as long as there is no contradiction between the combinations of the technical features, the scope of the present description should be considered.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is specific and detailed, but not to be understood as limiting the scope of the present invention. It should be noted that various changes and modifications can be made by those skilled in the art without departing from the spirit of the invention, and these changes and modifications are all within the scope of the invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A disc tracking method, comprising the steps of:
acquiring a collected near-infrared fundus video stream, and calculating the optic disc centroid position of each frame of fundus image in the near-infrared fundus video stream;
extracting initial front N frames of fundus images and the disc mass center position thereof in the near-infrared video stream, and determining an initial frame of a current judgment queue from the front N frames of fundus images according to the extracted disc mass center position; wherein the fixed length threshold of the current decision queue is n;
calculating the variance of the disc centroid position of the current frame fundus image and the disc centroid position of the previous effective frame fundus image and the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc for the current frame fundus image newly captured in the current determination queue; the distance between the optic disc mass center position of the current frame fundus image and the optic disc mass center position of the reference optic disc is a reference distance;
judging whether the optic disc centroid position of the current frame fundus image is a noise point or not according to the calculated variance and the reference distance; and when the optic disc center of mass of the current frame eye fundus image is judged to be a noise point, the current frame eye fundus image is removed; when the disc centroid position of the current frame fundus image is judged not to be a noise point, caching the current frame fundus image and the disc centroid position thereof;
continuously capturing the next frame of fundus images in the current judgment queue and judging until the length of the current judgment queue reaches a threshold value n, and eliminating the frame at the head of the queue according to a first-in first-out principle while entering the current judgment queue for each frame of fundus images captured subsequently.
2. The disc tracking method according to claim 1, wherein said extracting an initial first N frames of fundus images and the disc centroid position in the near-infrared video stream and determining an initial frame of a current decision queue from the first N frames of fundus images based on the extracted disc centroid position comprises the steps of:
extracting the first N frames of fundus images and the corresponding N video disk centroid positions from the near-infrared fundus video stream;
respectively carrying out average value calculation on the abscissa and the ordinate of the N video disc mass center positions to obtain an average mass center position;
calculating the distance between each of the mass center positions of the optic disc and the average mass center position to obtain N average distances;
sorting the N average distances in size, and removing a preset number of average distances in a descending order to obtain N-Q reasonable average distances; wherein Q is a preset number;
and selecting the maximum average distance from the reasonable average distances as a boundary value, comparing the maximum average distance with the boundary value as a reference according to the time sequence reverse order of the previous N frames, and acquiring a frame of which the distance between the first video disc centroid position and the average centroid position is less than or equal to the boundary value as the initial frame.
3. The disc tracking method according to claim 2, wherein the predetermined number Q has a value range of: q is more than or equal to N10% and less than or equal to N30%.
4. The disc tracking method according to claim 1, wherein said determining whether the disc centroid position of the current fundus image is noisy on the basis of the calculated variance and the reference distance comprises the steps of:
substituting the variance and the reference distance into a decision formula: p (m) a 1 ×(d m -d Datum )+a 2 ×l m Calculating to obtain a corresponding judgment value; wherein P (m) is a calculated judgment value, a 1 Is a coefficient of variance influence, a 2 As reference distance influence coefficient, d m Is the variance, d Reference(s) Is a reference variance,/ m Is the reference distance;
judging whether the judgment value is larger than or equal to a preset value;
when the judgment value is judged to be larger than or equal to the preset value, judging that the optic disc centroid position of the current fundus image is a noise point; and when the judgment value is judged to be smaller than the preset value, judging that the optic disc centroid position of the current fundus image is not a noise point.
5. The disc tracking method according to any one of claims 1 to 4, wherein the step of continuing to capture and determine the fundus image of the next frame further comprises the steps of:
counting the number of noise points judged currently;
and when counting that the centroid positions of the optic discs corresponding to the current continuous b frames of fundus images are all noise points, canceling the judgment of the next frame of fundus image and resetting the judgment queue.
6. A video disc tracking system is characterized by comprising a video disc mass center position calculating module, a judging queue determining module, a judging parameter calculating module and a noise point judging and processing module;
the optical disk centroid position calculation module is used for acquiring the acquired near-infrared fundus video stream and calculating the optical disk centroid position of each frame of fundus image in the near-infrared fundus video stream;
the judgment queue determining module is used for extracting initial front N frames of fundus images and the optic disc mass center position thereof in the near-infrared video stream, and determining the initial frame of the current judgment queue from the front N frames of fundus images according to the extracted optic disc mass center position; wherein the fixed length threshold of the current decision queue is n;
the judgment parameter calculation module is used for calculating the variance between the disc centroid position of the current frame fundus image and the disc centroid position of the previous effective frame fundus image and the distance between the disc centroid position of the current frame fundus image and the disc centroid position of the reference disc aiming at the current frame fundus image newly captured in the current judgment queue; the distance between the optic disc mass center position of the current frame fundus image and the optic disc mass center position of the reference optic disc is a reference distance;
the noise point judgment processing module is used for judging whether the centroid position of the optic disc of the current frame fundus image is a noise point according to the variance and the reference distance obtained by calculation; and when the optic disc center of mass of the current frame eye fundus image is judged to be a noise point, the current frame eye fundus image is removed; when the disc centroid position of the current frame fundus image is judged not to be a noise point, caching the current frame fundus image and the disc centroid position thereof to an output queue;
the noise point judgment processing module is further configured to skip to the noise point judgment module after the current frame fundus image is removed or cached, continue to capture and judge the next frame fundus image in the current judgment queue by the noise point judgment module until the length of the current judgment queue reaches the fixed threshold n, enter the current judgment queue for each frame fundus image captured subsequently, and simultaneously eliminate the frame at the head of the current judgment queue according to a first-in first-out principle.
7. The disc tracking system of claim 6, wherein the decision queue determining module includes an extraction sub-module, an average calculation sub-module, an average distance calculation sub-module, a sorting sub-module, and a selection comparison sub-module;
the extraction submodule is used for extracting all fundus images and the corresponding centroid positions of the optic discs from the near-infrared fundus video stream;
the average value calculation submodule is used for respectively carrying out average value calculation on the abscissa and the ordinate of the mass center positions of the first N optical disks to obtain an average mass center position;
the average distance calculation submodule is used for calculating the distance between each of the mass center positions of the optic disc and the average mass center position in the N optic disc mass center positions to obtain N average distances;
the sorting submodule is used for sorting the N average distances in size and eliminating the maximum average distances in a preset number from large to small to obtain N-Q reasonable average distances; wherein Q is a preset number;
and the selection and comparison submodule is used for selecting the maximum average distance from the reasonable average distances as a boundary value, comparing the maximum average distance with the boundary value as a reference according to the time sequence reverse order of the previous N frames, and acquiring a frame of which the distance between the first video disc centroid position and the average centroid position is smaller than or equal to the boundary value as the initial frame.
8. The disc tracking system of claim 6, wherein the noise judgment processing module includes a numerical substitution sub-module and a decision value judgment sub-module;
wherein the numerical value substitution submodule is configured to substitute the variance and the reference distance into a decision formula: p (m) ═ a 1 ×(d m -d Datum )+a 2 ×l m Calculating to obtain a corresponding judgment value; wherein P (m) is a calculated judgment value, a 1 Is a variance influence coefficient, a 2 As reference distance influence coefficient, d m Is the variance, d Datum Is a reference variance,/ m Is the reference distance;
the judgment value judgment submodule is used for judging whether the judgment value is larger than or equal to a preset value;
the judgment value judgment submodule is also used for judging that the position of the optic disc centroid of the current fundus image is a noise point when the judgment value is judged to be larger than or equal to the preset value; and when the judgment value is judged to be smaller than the preset value, judging that the optic disc centroid position of the current fundus image is not a noise point.
9. The disc tracking system of any of claims 6 to 8, further comprising a noise count module;
the noise number counting module is used for counting the number of currently determined noise in the judgment process that whether the optic disc centroid position of the next frame of fundus image in the current judgment queue is a noise or not by the noise judgment processing module;
the noise number counting module is further used for canceling the judgment of the next frame of fundus image when the disc mass center positions corresponding to the current continuous b frame of fundus image are counted to be noise, skipping to the judgment queue determining module, and resetting the judgment queue by the judgment queue determining module.
10. An eye fundus collection device comprising the disc tracking system of any one of claims 6 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811485787.XA CN109658395B (en) | 2018-12-06 | 2018-12-06 | Optic disc tracking method and system and eye fundus collection device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811485787.XA CN109658395B (en) | 2018-12-06 | 2018-12-06 | Optic disc tracking method and system and eye fundus collection device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109658395A CN109658395A (en) | 2019-04-19 |
| CN109658395B true CN109658395B (en) | 2022-09-09 |
Family
ID=66112351
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811485787.XA Active CN109658395B (en) | 2018-12-06 | 2018-12-06 | Optic disc tracking method and system and eye fundus collection device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109658395B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117788407A (en) * | 2019-12-04 | 2024-03-29 | 深圳硅基智能科技有限公司 | Training method for glaucoma image feature extraction based on artificial neural network |
| CN112686927B (en) * | 2020-12-31 | 2023-05-12 | 上海易维视科技有限公司 | Human eye position regression calculation method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107180421A (en) * | 2016-03-09 | 2017-09-19 | 中兴通讯股份有限公司 | A kind of eye fundus image lesion detection method and device |
| CN107424128A (en) * | 2017-06-20 | 2017-12-01 | 南京泰立瑞信息科技有限公司 | A kind of method and system that picture rich in detail frame is extracted from the fundus video recorded |
| CN108024707A (en) * | 2015-07-24 | 2018-05-11 | 伟伦公司 | Automatic eye fundus image capture systems |
| WO2018116321A2 (en) * | 2016-12-21 | 2018-06-28 | Braviithi Technologies Private Limited | Retinal fundus image processing method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6215555B2 (en) * | 2013-03-29 | 2017-10-18 | 株式会社ニデック | Fundus image processing apparatus and fundus image processing program |
-
2018
- 2018-12-06 CN CN201811485787.XA patent/CN109658395B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108024707A (en) * | 2015-07-24 | 2018-05-11 | 伟伦公司 | Automatic eye fundus image capture systems |
| CN107180421A (en) * | 2016-03-09 | 2017-09-19 | 中兴通讯股份有限公司 | A kind of eye fundus image lesion detection method and device |
| WO2018116321A2 (en) * | 2016-12-21 | 2018-06-28 | Braviithi Technologies Private Limited | Retinal fundus image processing method |
| CN107424128A (en) * | 2017-06-20 | 2017-12-01 | 南京泰立瑞信息科技有限公司 | A kind of method and system that picture rich in detail frame is extracted from the fundus video recorded |
Non-Patent Citations (2)
| Title |
|---|
| 基于置信度计算的快速眼底图像视盘定位;吴慧等;《计算机辅助设计与图形学学报》;20170615(第06期);全文 * |
| 彩色眼底图像视盘自动定位与分割;邹北骥等;《光学精密工程》;20150415(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109658395A (en) | 2019-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9819864B2 (en) | Wide field retinal image capture system and method | |
| CN104036231B (en) | Trunk identification device and method, terminal image detection method, device | |
| CN109145803B (en) | Gesture recognition method and device, electronic equipment and computer readable storage medium | |
| WO2019001505A1 (en) | Target feature extraction method and device, and application system | |
| CN109658395B (en) | Optic disc tracking method and system and eye fundus collection device | |
| CN110264493A (en) | A kind of multiple target object tracking method and device under motion state | |
| KR101051569B1 (en) | Image Noise Reduction Device and Method | |
| CN109547677B (en) | Fundus image shooting method, system and equipment | |
| CN104135612B (en) | A kind of image pickup method and filming apparatus of adjustable shooting body position | |
| CN109508648A (en) | A kind of face snap method and apparatus | |
| CN109886995B (en) | Multi-target tracking method in complex environment | |
| CN110796678A (en) | IoU-based underwater multi-target tracking method | |
| CN113436070B (en) | Fundus image splicing method based on deep neural network | |
| KR101064927B1 (en) | People count system using image analysis and method thereof | |
| CN114531549B (en) | Image acquisition method, electronic device, and computer-readable storage medium | |
| CN111986163A (en) | Face image selection method and device | |
| CN111444555B (en) | Temperature measurement information display method and device and terminal equipment | |
| CN113822927B (en) | Face detection method, device, medium and equipment suitable for weak quality image | |
| TWI583192B (en) | Image processing device | |
| RU2616152C1 (en) | Method of spatial position control of the participants of the sports event on the game field | |
| CN111353326A (en) | In-vivo detection method based on multispectral face difference image | |
| CN117392404A (en) | Method and system for improving image detection speed | |
| CN112861588B (en) | Living body detection method and device | |
| CN114283356B (en) | Acquisition and analysis system and method for moving image | |
| RU2542876C2 (en) | Apparatus for selecting highly detailed objects on scene image |
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: 20231220 Address after: Room 707, Unit 1, Building 1, Courtyard 1, Longyu Middle Street, Huilongguan Town, Changping District, Beijing, 100085 Patentee after: Beijing Zhizhen Health Technology Co.,Ltd. Address before: 100085 room 1210, Building 29, yard 9, anningzhuang West Road, Haidian District, Beijing Patentee before: Dai Liming |
|
| TR01 | Transfer of patent right |