CN106447694A - Video badminton motion detection and tracking method - Google Patents
Video badminton motion detection and tracking method Download PDFInfo
- Publication number
- CN106447694A CN106447694A CN201610602578.3A CN201610602578A CN106447694A CN 106447694 A CN106447694 A CN 106447694A CN 201610602578 A CN201610602578 A CN 201610602578A CN 106447694 A CN106447694 A CN 106447694A
- Authority
- CN
- China
- Prior art keywords
- badminton
- video
- target
- shuttlecock
- histogram
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30221—Sports video; Sports image
- G06T2207/30224—Ball; Puck
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a video badminton motion detection and tracking method. According to the method, foreground motion objects are extracted by utilizing a wavelet multi-scale characteristic based on a video inter-frame difference according to a condition that wavelet transform has excellent local characteristics in a time domain and a space domain; the foreground motion objects in a video scene are classified and detected through sample learning and training of positive and negative badminton targets; a badminton target is determined; and a badminton is tracked by adopting a particle filter based on difference of color features of the badminton. Specific hardware support and scene condition constraints are not needed, so that the method is simple and convenient, and is flexible and easy to realize.
Description
Technical field
The present invention relates to a kind of video badminton detection and tracking, be used for digital image analysis and understanding.Belong to
In intelligent information processing technology field.
Background technology
With improving constantly of athletics sports technical merit, day by day require science, perfect training method and high efficiency smart
Technical-tactics analyzing and decision-making mechanism improving athletics sports technical merit, and badminton as one have high antagonism,
Quickly, the athletic sports project of the feature such as flexibly, more needs the skill of science, perfect training method and high efficiency smart to fight
Art is analyzed and decision-making mechanism, requires to meet its flexible, complicated technique and tactics system.Wherein, in badminton game and training,
Owing to shuttlecock flying speed is fast, circuit and drop point change are various, therefore, how to improve shuttlecock technical and tactical levels and achievement,
Challenge is how the detection of effective intelligent online and the badminton position followed the tracks of in video scene.
Owing to shuttlecock metamorphosis is various, flying speed is fast, and circuit and drop point change are various, and video scene change is multiple
Miscellaneous various, the video badminton detection causing online effectively intelligence is very difficult with tracking.Currently mainly use supervised
Manual method carry out detection and the positioning of badminton target, therefore, be easily caused the detection of badminton target with
Positioning result is affected by artificial subjective factor, and working strength is high, wastes time and energy.
Content of the invention
Present invention aims to existing video badminton detection can be low with tracking reliability in time, inspection
Surveying and following the tracks of result by dynamic scene sensitive, noise jamming greatly, difficulty meets badminton technique and tactics and analyzes requirement in time,
A kind of video badminton detection and tracking are provided, by sample learning and the training of target positive and negative to shuttlecock, right
Foreground moving object in video scene carries out classifying and detects, and determines shuttlecock target, and based on shuttlecock color character
Otherness, uses particle filter, it is achieved effective tracking of the badminton target under the conditions of multiple.
For reaching above-mentioned purpose, idea of the invention is that:It is respectively provided with excellent office according to wavelet transformation in time domain and spatial domain
Portion's feature, based on video frame-to-frame differences, utilizes multi-scale wavelet characteristic, extracts foreground moving object, by positive and negative to shuttlecock
The sample learning of target and training, carry out classifying to the foreground moving object in video scene and detect, determining shuttlecock target,
And based on the otherness of shuttlecock color character, use particle filter, shuttlecock is tracked.
Conceiving according to foregoing invention, the present invention uses following technical proposals:
A kind of video badminton detection and tracking, use and connected the shuttlecock fortune that computer is constituted by video camera
Dynamic detection operates with following the tracks of image capturing system, it is characterised in that comprise the following steps that:
1) start badminton detection and follow the tracks of image capturing system:Gather video image;
2) foreground moving object segmentation
By the current frame image of camera acquisition and previous frame image subtraction, small wave converting method is used to be partitioned into prospect fortune
Dynamic subject area;
3) sample learning and training;
4) badminton target detection;
5) badminton target following.
Above-mentioned steps 2) concrete operation step as follows:
(1) current frame image It(x, y) with previous frame image It-1(x, y) subtracts each other, obtain difference image D (x, y):
D (x, y)=It(x,y)-It-1(x,y);
(2) difference image multi-scale wavelet transformation:
Wherein, D is difference image, and h, v are respectively level, the filter operator in vertical direction,For convolution;
(3) determination in foreground moving object region:Determine threshold value T of difference image multi-scale wavelet transformation E1, E value is high
In T1All pixels composition region, be defined as foreground moving object region.
Above-mentioned steps 3) concrete operation step as follows:
(1) according to step 2), gather the shuttlecock Haar feature under different video scene, constitute shuttlecock training sample
Data acquisition system Di={ Hi, and the Haar feature of non-shuttlecock classification, constitute the tag set C of non-shuttlecocki={ Ti};
(2) selection sort device, to above-mentioned data acquisition system DiWith tag set CiSample set (the D constitutingi, Ci) supervise
Educational inspector practises, and adjusts parameter in grader, makes classifying quality reach optimal.
Above-mentioned steps 4) concrete operation step as follows:
(1) according to step 2), gather the Haar feature of foreground moving object, constitute test data set and close ADi={ AHi};
(2) according to step 3) determined by grader and parameter thereof, to test data set close ADiCarry out discriminant classification, really
Determine badminton target.
Above-mentioned steps 5) concrete operation step as follows:
(1) color space conversion:By the red R of RGB color space, green G, blue B three-component, determine the look of HSV color space
Adjust component H, saturation degree component S and luminance component V:
Wherein,
V=max (R, G, B)
(2) shuttlecock feature histogram builds:According to step 4) determined by badminton target, use HSV color
Chrominance component H in space, saturation degree component S, set up the color histogram of each component m levelAnd use HSV color space
In luminance component V, set up n level gray gradient histogramAnd then set up confluent colours (H, S) histogramAnd brightness
(V) gray gradient histogramBadminton target signature histogram qr:
Wherein, C is normalization coefficient,
(3) badminton target following:The badminton target signature histogram q building according to step (2)r, adopt
With particle filter, the badminton target in video scene is tracked.
The principle of the present invention is as follows:In the inventive solutions, it is respectively provided with in time domain and spatial domain according to wavelet transformation
Excellent local characteristic, based on video frame-to-frame differences, utilizes multi-scale wavelet characteristic, extracts foreground moving object, by plumage
The sample learning of the positive and negative target of ball top and training, carry out classifying to the foreground moving object in video scene and detect, determining plumage
Ball top target, and based on the otherness of shuttlecock color character, use particle filter, shuttlecock is tracked.
If a certain moment, obtain adjacent two two field picture f (t respectivelyn-1, x, y), f (tn, x, y), by two width images by pixel
Seek difference, obtain difference image Diff (x, y):
DiffR (x, y)=| fR (tn,x,y)-fR(tn-1,x,y)|
DiffG (x, y)=| fG (tn,x,y)-fG(tn-1,x,y)|
DiffB (x, y)=| fB (tn,x,y)-fB(tn-1,x,y)|
Wherein, DiffR, DiffG, DiffB corresponding difference image red, green, blue three-component respectively, the absolute value that | f | is f.
Based on above-mentioned neighbor frame difference, use wavelet transformation, be partitioned into foreground moving object region.According to two-dimensional image I (x,
Y) at yardstick 2jWith the wavelet transformation on k direction:
Then at x, the wavelet function on y direction is represented by:
In formula, (x y) is smoothing filter function to θ.
Thus can determine that (x, y) (x, y) after smothing filtering, the wavelet transformation under different scale is image I through function #:
If gradient amplitudeReach local maximum along following gradient direction, then in image this point (x y) is multiple dimensioned limit
Edge point
Accordingly, it may be determined that the marginal point under different scale.Owing to noise is sensitive to dimensional variation, therefore, use above-mentioned seeking
Seek local amplitude maximum, it is impossible to effectively compacting noise.It for effectively overcoming this impact, is higher than certain threshold by seeking gradient amplitude
Value method, substitutes and seeks local amplitude maximum, determine the marginal point of different scale.
Wherein, h, v are respectively level, the filter operator in vertical direction, T1For threshold value,For convolution operator.
Consider shuttlecock feature mode space X, comprise m pattern xiTraining set and corresponding class label ωi, and
It is assumed to two class classification problems.In each layer of k, the importance of sample uses weight set DkI () reflects, and meet
In two classification problems, the study of Weak Classifier makes object function εkMinimize:
Wherein, P [.] is the empirical probability based on training sample observation.
Update weight D as the following formulak(i):
Wherein, ZkFor normalization factor, meet
Final grader is being considered its weight α by all k Weak ClassifierskRear weight a majority of the votes cast determines.
Based on shuttlecock target location determined by above-mentioned grader, use the chrominance component H in HSV color space, satisfy
With degree component S, set up the color histogram of each component m levelAnd use luminance component V in HSV color space, set up n
Level gray gradient histogramOn this basis, confluent colours (H, S) histogram is set upStraight with brightness (V) shade of gray
Fang TuShuttlecock feature histogram qr:
Wherein, C is normalization coefficient,
If Xt, ZtIt is respectively shuttlecock dbjective state and the observation of t, then shuttlecock tracking problem is converted into and asks
Solve posterior probability p (Xt|Z1:t), wherein, Z1:t=(Z1,…,Zt) by the shuttlecock target observation value being obtained to t.
Use one group of particle with weightClose approximation posterior probability p (Xt|Z1:t), wherein,
For particle, the shuttlecock dbjective state expressing possibility,Weight for particle.
New particle is produced by resampling function, and this functional dependence is in shuttlecock dbjective state and observation, i.e.
Following weight is used to be updated new particle:
And new particle is produced by following state transition function:
Xt=Ft(Xt-1, Ut)
Wherein, UtFor system noise, FtMotion state for shuttlecock target.
The present invention compared with prior art, has following obvious prominent substantive distinguishing features and remarkable advantage:This
Invention is respectively provided with excellent local characteristic according to wavelet transformation in time domain and spatial domain, based on video frame-to-frame differences, utilizes small echo many
Dimensional properties, extracts foreground moving object, by sample learning and the training of target positive and negative to shuttlecock, in video scene
Foreground moving object carry out classifying and detect, determine shuttlecock target, and based on the otherness of shuttlecock color character, use
Particle filter, is tracked to shuttlecock, simple operation, flexibly, easily realizes, solve video badminton detection with
During tracking, it is desirable to heavy hand labor, and it is single with tracking object to limit detection, does dynamic scene sensitive, noise
Disturb big, computing complexity, and specific hardware supported and scene condition constraint;Improve video badminton detection with
The robustness of track, the video badminton being suitable under various complex background condition detection and tracking.
Brief description
Fig. 1 is the flowsheet of the inventive method.
Fig. 2 is the original current frame image of video of one embodiment of the invention.
Fig. 3 is the two-value foreground moving object area image being partitioned in Fig. 2 example.
Fig. 4 is the foreground moving object area image being partitioned in Fig. 2 example.
Fig. 5 is the shuttlecock testing result (rectangular box) in Fig. 4 example.
Fig. 6 is that result (rectangular box) followed the tracks of by the shuttlecock in Fig. 2 example.
Detailed description of the invention
It is as follows that the preferred embodiments of the present invention combine detailed description:
Implement one:
See Fig. 1~Fig. 6, the badminton detection of this video and tracking, use and video camera is connected calculating mechanism
The badminton detection becoming operates with following the tracks of image capturing system, it is characterised in that comprise the following steps that:
1) start badminton detection and follow the tracks of image capturing system:Gather video image;
2) foreground moving object segmentation
By the current frame image of camera acquisition and previous frame image subtraction, use small wave converting method, be partitioned into prospect
Moving Objects region;
3) sample learning and training;
4) badminton target detection;
5) badminton target following.
Embodiment two:
This preferred embodiment is:The operation sequence of video badminton detection and tracking is as shown in Figure 1.This example
Original current frame image is as in figure 2 it is shown, the image shown in Fig. 2 carries out consecutive frame difference, and carries out multi-scale wavelet transform, enters
Row foreground moving object is split, and obtains two-value foreground moving object region as it is shown on figure 3, by target positive and negative to shuttlecock
Sample learning and training, carry out classifying to the foreground moving object in video scene and detect, and determines shuttlecock target, and based on
The otherness of shuttlecock color character, uses particle filter, is tracked shuttlecock;Concrete operation step is as follows:
1) start badminton detection and follow the tracks of image capturing system:Gather video image;
2) foreground moving object segmentation:Concrete operation step is as follows:
(1) by camera acquisition such as the current frame image I of Fig. 21(x, y) with previous frame image I2(x, y) subtracts each other, and obtains
Difference image D (x, y):
D (x, y)=It(x,y)-It-1(x,y);
(2) difference image multi-scale wavelet transformation:
Wherein, D is difference image, and h, v are respectively level, the filter operator in vertical direction,For convolution;
(3) determination in foreground moving object region:Determine threshold value T of difference image multi-scale wavelet transformation E1, E value is high
In T1All pixels composition region, be defined as foreground moving object region.
Fig. 3 is the two-value foreground moving object region through above-mentioned gained, and Fig. 4 is the foreground moving object being partitioned into.
3) sample learning and training:Gather the shuttlecock Haar feature under different video scene, constitute shuttlecock training sample
This data acquisition system Di={ Hi, and the Haar feature of non-shuttlecock classification, constitute the tag set C of non-shuttlecocki=
{Ti, use SVMs and select Radial basis kernel function, to above-mentioned data acquisition system DiWith tag set CiThe sample set constituting
Close (Di, Ci) carry out learning and training, constantly penalty factor parameter γ in modification Radial basis kernel function, makes correct recognition rata reach
Arrive the highest.
4) badminton target detection:Foreground moving object shown in Fig. 3, gathers Haar feature, constitutes test number
According to set ADi={ AHi, use fixed penalty factor parameter γ to carry out dividing based on the SVMs of Radial basis kernel function
Class differentiates, determines badminton target.Rectangular box in Fig. 5 show the shuttlecock position through above-mentioned gained.
5) badminton target following
Concrete operation step is as follows:
(1) color space conversion:By the red R of RGB color space, green G, blue B three-component, determine the look of HSV color space
Adjust component H, saturation degree component S and luminance component V:
Wherein,
V=max (R, G, B)
(2) shuttlecock feature histogram builds:Shuttlecock target to Fig. 5 example, uses the tone in HSV color space
Component H, saturation degree component S, set up the color histogram of each 8 grades of H and S componentAnd use brightness in HSV color space
Component V, sets up 8 grades of gray gradient histogramsAnd then set up confluent colours (H, S) histogramWith brightness (V) shade of gray
HistogramShuttlecock feature histogram qr:
Wherein, C is normalization coefficient,
(3) shuttlecock target following:According to constructed shuttlecock feature histogram qr, use particle filter, to field
Shuttlecock target in scape is tracked.Rectangular box in Fig. 6 show follows the tracks of result through the shuttlecock of above-mentioned gained.
Claims (5)
1. a video badminton detection and tracking, it is characterised in that comprise the following steps that:
1) start badminton detection and follow the tracks of image capturing system:Gather video image;
2) foreground moving object segmentation
By the current frame image of camera acquisition and previous frame image subtraction, use small wave converting method, be partitioned into foreground moving
Subject area;
3) sample learning and training;
4) badminton target detection;
5) badminton target following.
2. video badminton detection according to claim 1 and tracking, it is characterised in that described step 2) front
The concrete operation step of scape motion segmentation is as follows:
(1) current frame image It(x, y) with previous frame image It-1(x, y) subtracts each other, obtain difference image D (x, y):
D (x, y)=It(x,y)-It-1(x,y);
(2) difference image multi-scale wavelet transformation:
Wherein, D is difference image, and h, v are respectively level, the filter operator in vertical direction,For convolution;
(3) determination in foreground moving object region:Determine threshold value T of difference image multi-scale wavelet transformation E1, E value is higher than T1
All pixels composition region, be defined as foreground moving object region.
3. video badminton detection according to claim 1 and tracking, it is characterised in that described step 3) sample
This study is as follows with the concrete operation step of training:
(1) according to step 2), gather the shuttlecock Haar feature under different video scene, constitute the data of shuttlecock training sample
Set Di={ Hi, and the Haar feature of non-shuttlecock classification, constitute the tag set C of non-shuttlecocki={ Ti};
(2) selection sort device, to above-mentioned data acquisition system DiWith tag set CiSample set (the D constitutingi, Ci) exercising supervision
Practise, and adjust parameter in grader, make classifying quality reach optimal.
4. video badminton detection according to claim 1 and tracking, it is characterised in that described step 4) plumage
The concrete operation step of ball top moving object detection is as follows:
(1) according to step 2), gather the Haar feature of foreground moving object, constitute test data set and close ADi={ AHi};
(2) according to step 3) determined by grader and parameter thereof, to test data set close ADiCarry out discriminant classification, determine feather
Ball moving target.
5. video badminton detection according to claim 1 and tracking, it is characterised in that described step 5) plumage
The concrete operation step of ball top motion target tracking is as follows:
(1) color space conversion:By the red R of RGB color space, green G, blue B three-component, determine that the tone of HSV color space divides
Amount H, saturation degree component S and luminance component V:
Wherein,
V=max (R, G, B)
(2) shuttlecock feature histogram builds:According to step 4) determined by badminton target, use HSV color space
In chrominance component H, saturation degree component S, set up the color histogram of each component m levelAnd use in HSV color space
Luminance component V, sets up n level gray gradient histogramAnd then set up confluent colours (H, S) histogramWith brightness (V) ash
Degree histogram of gradientsBadminton target signature histogram qr:
Wherein, C is normalization coefficient,
(3) badminton target following:The badminton target signature histogram q building according to step (2)r, use particle
Wave filter, is tracked to the badminton target in video scene.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610602578.3A CN106447694A (en) | 2016-07-28 | 2016-07-28 | Video badminton motion detection and tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610602578.3A CN106447694A (en) | 2016-07-28 | 2016-07-28 | Video badminton motion detection and tracking method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106447694A true CN106447694A (en) | 2017-02-22 |
Family
ID=58184299
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610602578.3A Pending CN106447694A (en) | 2016-07-28 | 2016-07-28 | Video badminton motion detection and tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106447694A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107895365A (en) * | 2017-11-01 | 2018-04-10 | 国网山东省电力公司电力科学研究院 | The image matching method and monitoring system of broken protection outside a kind of passway for transmitting electricity |
CN108009504A (en) * | 2017-12-04 | 2018-05-08 | 深圳市赢世体育科技有限公司 | A kind of recognition methods of moving sphere, device and storage medium |
CN108036957A (en) * | 2017-11-28 | 2018-05-15 | 安徽省蓝翔体育用品有限公司 | A kind of shuttlecock cuts mould detection method |
CN108079525A (en) * | 2017-11-28 | 2018-05-29 | 安徽省蓝翔体育用品有限公司 | A kind of feather shape selection system based on the shuttlecock service life |
CN108168852A (en) * | 2017-11-28 | 2018-06-15 | 安徽省蓝翔体育用品有限公司 | A kind of shuttlecock lashing wire position measuring system |
CN113379702A (en) * | 2021-06-08 | 2021-09-10 | 广州医软智能科技有限公司 | Blood vessel path extraction method and device of microcirculation image |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739689A (en) * | 2009-12-15 | 2010-06-16 | 中国科学院长春光学精密机械与物理研究所 | Real-time extraction and update method of background image in motion video |
CN102968802A (en) * | 2012-11-28 | 2013-03-13 | 无锡港湾网络科技有限公司 | Moving target analyzing and tracking method and system based on video monitoring |
CN103325112A (en) * | 2013-06-07 | 2013-09-25 | 中国民航大学 | Quick detecting method for moving objects in dynamic scene |
CN104036250A (en) * | 2014-06-16 | 2014-09-10 | 上海大学 | Video pedestrian detecting and tracking method |
CN104874164A (en) * | 2015-05-06 | 2015-09-02 | 芜湖林一电子科技有限公司 | Visual tracking badminton training partner robot |
CN105477845A (en) * | 2015-06-17 | 2016-04-13 | 电子科技大学 | Method and system for predicting motion trajectory of shuttlecock |
-
2016
- 2016-07-28 CN CN201610602578.3A patent/CN106447694A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739689A (en) * | 2009-12-15 | 2010-06-16 | 中国科学院长春光学精密机械与物理研究所 | Real-time extraction and update method of background image in motion video |
CN102968802A (en) * | 2012-11-28 | 2013-03-13 | 无锡港湾网络科技有限公司 | Moving target analyzing and tracking method and system based on video monitoring |
CN103325112A (en) * | 2013-06-07 | 2013-09-25 | 中国民航大学 | Quick detecting method for moving objects in dynamic scene |
CN104036250A (en) * | 2014-06-16 | 2014-09-10 | 上海大学 | Video pedestrian detecting and tracking method |
CN104874164A (en) * | 2015-05-06 | 2015-09-02 | 芜湖林一电子科技有限公司 | Visual tracking badminton training partner robot |
CN105477845A (en) * | 2015-06-17 | 2016-04-13 | 电子科技大学 | Method and system for predicting motion trajectory of shuttlecock |
Non-Patent Citations (4)
Title |
---|
RUIYUE XU ET AL: "Multiple human detection and tracking based on head detection for real-time video surveillance", 《MULTIMEDIA TOOLS AND APPLICATIONS》 * |
张晶华 等: "具有弯拱度羽毛片的特性及其应用", 《广东工业大学学报》 * |
胡威: "一种新的Meanshift 快速运动目标鲁棒跟踪方法", 《计算机应用与软件》 * |
苏振阳 等: "基于机器视觉的羽毛球线审辅助系统研究", 《电子世界》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107895365A (en) * | 2017-11-01 | 2018-04-10 | 国网山东省电力公司电力科学研究院 | The image matching method and monitoring system of broken protection outside a kind of passway for transmitting electricity |
CN107895365B (en) * | 2017-11-01 | 2020-08-18 | 国网山东省电力公司电力科学研究院 | Image matching method and monitoring system for power transmission channel external damage protection |
CN108036957A (en) * | 2017-11-28 | 2018-05-15 | 安徽省蓝翔体育用品有限公司 | A kind of shuttlecock cuts mould detection method |
CN108079525A (en) * | 2017-11-28 | 2018-05-29 | 安徽省蓝翔体育用品有限公司 | A kind of feather shape selection system based on the shuttlecock service life |
CN108168852A (en) * | 2017-11-28 | 2018-06-15 | 安徽省蓝翔体育用品有限公司 | A kind of shuttlecock lashing wire position measuring system |
CN108009504A (en) * | 2017-12-04 | 2018-05-08 | 深圳市赢世体育科技有限公司 | A kind of recognition methods of moving sphere, device and storage medium |
CN108009504B (en) * | 2017-12-04 | 2020-06-26 | 深圳市赢世体育科技有限公司 | Moving sphere identification method and device and storage medium |
CN113379702A (en) * | 2021-06-08 | 2021-09-10 | 广州医软智能科技有限公司 | Blood vessel path extraction method and device of microcirculation image |
CN113379702B (en) * | 2021-06-08 | 2024-05-28 | 广州医软智能科技有限公司 | Blood vessel path extraction method and device for microcirculation image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106447694A (en) | Video badminton motion detection and tracking method | |
CN102214298B (en) | Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism | |
Shao et al. | Car detection from high-resolution aerial imagery using multiple features | |
CN102722891B (en) | Method for detecting image significance | |
CN104036250B (en) | Video pedestrian detection and tracking | |
CN102414720B (en) | Characteristic quantity calculation element, characteristic quantity calculating method | |
CN107909081B (en) | Method for quickly acquiring and quickly calibrating image data set in deep learning | |
CN108549864A (en) | Area-of-interest filter method towards vehicle-mounted thermal imaging pedestrian detection and device | |
KR20160143494A (en) | Saliency information acquisition apparatus and saliency information acquisition method | |
CN104978567B (en) | Vehicle checking method based on scene classification | |
CN107491762A (en) | A kind of pedestrian detection method | |
CN108257151B (en) | PCANet image change detection method based on significance analysis | |
CN103413120A (en) | Tracking method based on integral and partial recognition of object | |
CN102842032A (en) | Method for recognizing pornography images on mobile Internet based on multi-mode combinational strategy | |
CN106650668A (en) | Method and system for detecting movable target object in real time | |
CN106203237A (en) | The recognition methods of container-trailer numbering and device | |
CN109918971A (en) | Number detection method and device in monitor video | |
CN106709438A (en) | Method for collecting statistics of number of people based on video conference | |
CN106599880A (en) | Discrimination method of the same person facing examination without monitor | |
CN109190456A (en) | Pedestrian detection method is overlooked based on the multiple features fusion of converging channels feature and gray level co-occurrence matrixes | |
Manafifard et al. | Multi-player detection in soccer broadcast videos using a blob-guided particle swarm optimization method | |
Ma et al. | An all-weather lane detection system based on simulation interaction platform | |
CN104680189B (en) | Based on the bad image detecting method for improving bag of words | |
CN107330918B (en) | Football video player tracking method based on online multi-instance learning | |
CN108765463A (en) | A kind of moving target detecting method calmodulin binding domain CaM extraction and improve textural characteristics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170222 |
|
RJ01 | Rejection of invention patent application after publication |