CN105335711B - Fingertip Detection under a kind of complex environment - Google Patents
Fingertip Detection under a kind of complex environment Download PDFInfo
- Publication number
- CN105335711B CN105335711B CN201510700440.2A CN201510700440A CN105335711B CN 105335711 B CN105335711 B CN 105335711B CN 201510700440 A CN201510700440 A CN 201510700440A CN 105335711 B CN105335711 B CN 105335711B
- Authority
- CN
- China
- Prior art keywords
- finger tip
- mass center
- hand region
- away
- circle
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/107—Static hand or arm
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides Fingertip Detection under a kind of complex environment, comprising: the first step passes through and calculates dense optical flow information corresponding with scene information；Complexion filter device is constructed again obtains hand region；Second step is constructed the model of each posture hand region using homalographic block, calculates the mass center of hand region；Calculate all configuration sampling points to mass center distance peace mass center away from；Determined extended centroid away from using mass center as the center of circle, extended centroid is round away from picture is carried out for radius according to the finger tip number having detected that；Reject the profile point and the upper the largest number of wrist areas of contiguous pixels of circle in circle, circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；The number that this wheel is detected finger tip judges whether finger tip continues to test compared with the last round of finger tip number having detected that.This method strong robustness can also correctly detect finger tip when manpower is freely moved before camera under complex environment, to improve the accuracy and validity of finger tip detection.
Description
Technical field
The present invention relates to image processing and analysis technical fields, examine more specifically to finger tip under a kind of complex environment
Survey method.
Background technique
Traditional man-machine interactive system mainly passes through the information that the media such as button, mouse, keyboard carry out the mankind and computer
There is the disadvantages of function is limited, occupied space is big mostly, constrain the development of human-computer interaction technology in exchange, these interactive modes.
In recent years, with the development of computer science and artificial intelligence technology, human-computer interaction is gradually intended to nature, intuitive way,
Wherein, the spies such as the man-machine interaction mode based on computer vision technique is easy to use by its, flexibility is high, noise resisting ability is strong
Point has obtained the concern of more and more researchers.
As the important component of the man-machine interactive system based on computer vision technique, finger tip detection is known in gesture
Not, the fields such as virtual controlling have very wide application prospect.The finger tip detection of early stage is marked or is worn by special pigment
The modes such as LED light detect finger tip from scene, although this method is simple, use very inconvenient, are only used for certain
A little simple scenes；Thereafter, the use of wearable data glove improves the robustness of finger tip detection to a certain extent, still
Its convenience is poor, while wearable data glove price general charged is higher；With the development of camera technology, researcher
Begin to use some special cameras (such as Kinect) to obtain special scene information, is obviously mentioned although finger tip detection effect has
Height, but such camera popularization is not high, it is difficult to it promotes.At the same time, many researchers propose using common
Camera carries out the algorithm of finger tip detection, but these algorithms are mostly fairly simple, and manpower inspection can not be carried out under complex environment
It surveys.In conclusion currently employed common camera capturing scenes information carries out finger tip detection still man-machine hand interactive system
Research tendency.
Finger tip detection technology still remains many problems and not yet solves at present: (1) manpower is non-rigid object, is had very high
Freedom degree, therefore the finger under different situations can not be matched with fixed template；(2) the finger appearance of different people exists poor
Not, the finger shape and posture of different people are accurately detected out, there is larger difficulty；(3) even the finger of the same person,
There is also fine distinctions, if these nuances can not be identified correctly, will bring large error to testing result；(4) from
Dividing hand region in scene is to carry out the basis of finger tip detection, and when environment is complex, the Accurate Segmentation of hand region has
There is larger difficulty.
Summary of the invention
It is an object of the invention to overcome shortcoming and deficiency in the prior art, finger tip detection under a kind of complex environment is provided
Method；The Fingertip Detection not only can effectively realize the segmentation that hand region is carried out under complex environment, but also the detection side
Method strong robustness can also correctly detect finger tip when manpower is freely moved before camera under complex environment, refer to improve
The accuracy and validity of point detection.
In order to achieve the above object, the technical scheme is that: finger tip is examined under a kind of complex environment
Survey method is detected for the finger tip to hand；It is characterized by comprising following two steps:
The first step, capturing scenes information, by calculating dense optical flow information corresponding with scene information and carrying out binaryzation
Pretreatment obtains the scene image containing hand region；Construct Complexion filter device again, to the scene image containing hand region into
The segmentation of row hand region, obtains hand region；
Second step constructs the model of each posture hand region using homalographic block, according to the areal calculation hand of hand region
The mass center in portion region；Hand region contour is sampled again, all configuration sampling points is calculated to the distance of mass center, and calculates
Average mass center away from；Then determined extended centroid away from expanding and using mass center as the center of circle according to the last round of finger tip number having detected that
Exhibition mass center is round away from picture is carried out for radius；The profile point and the upper the largest number of wrist areas of contiguous pixels of circle in circle are finally rejected,
And circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；This wheel is detected to the number of finger tip
Compared with the last round of finger tip number having detected that, if the two is equal, realizes the detection of finger tip, otherwise detect this wheel
The number of finger tip is updated to the finger tip number having detected that, come determine extended centroid away from and carry out picture circle, continue the detection of finger tip.
In the first step, the capturing scenes information is gone forward side by side by calculating dense optical flow information corresponding with scene information
Row binaryzation pre-processes to obtain the scene image containing hand region；Complexion filter device is constructed again, to the field containing hand region
Scape image carries out the segmentation of hand region, obtains hand region and refers to: the following steps are included:
(1.1) capturing scenes information, and calculate corresponding dense optical flow information；
(1.2) Optic flow information is traversed, maximum light stream value is found, and utilize the maximum light stream value, in X-axis and Y direction pair
All light streams carry out Regularization；
(1.3) according to the light stream after regularization, tone and saturation degree along the changed region in light stream direction are calculated, and
Mark different color values；
(1.4) given threshold converts bianry image for the light stream region of variation of color mark according to threshold value, in conjunction with logic
Operation and mathematical morphological operation obtain the scene image containing hand region.
(1.5) according to human body complexion Clustering features, YCbCr Complexion filter device is constructed, rejects redundancy color and luminance information；
(1.6) all profiles of the scene image containing hand region are passed after being filtered to YCbCr Complexion filter device
Emission reduction sequence finds largest connected region as hand region, realizes and carry out hand region to the scene image containing hand region
Segmentation.
In second step, the model that each posture hand region is constructed using homalographic block, to determine hand region
Area, to calculate the mass center of hand region；Hand region contour is sampled again, calculates all configuration sampling points to mass center
Distance, and calculate average mass center away from；Then determined according to the last round of finger tip number having detected that extended centroid away from, and with
Mass center is the center of circle, and extended centroid is round away from picture is carried out for radius；The upper contiguous pixels number of profile point and circle in finally rejecting is round is most
More wrist areas, and circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；This wheel is detected
To finger tip number compared with the last round of finger tip number having detected that, if the two is equal, realize the detection of finger tip, otherwise will
This wheel detects the finger tip number that the number of finger tip is updated to have detected that, come determine extended centroid away from and carry out picture circle, after
The detection of continuous finger tip refers to: the following steps are included:
(2.1) homalographic is obtained according to the different postures of hand region with homalographic block building hand region display model
The quantity of block, to determine the area of hand region；
(2.2) mass center of hand region is calculated according to the coordinate of the area of hand region and each homalographic block；
(2.3) intensive sampling is carried out to hand region contour, calculates the profile point of all samplings to the distance of mass center, and count
Calculate average mass center away from；
(2.4) determine extended centroid away from D according to the finger tip number N having detected thatext, and using mass center as the center of circle, extended centroid
Away from DextPicture circle is carried out for radius；And rejecting includes the profile point in the circle；Wherein, the initial value of N is set as 0；
(2.5) number of pixels of the circle Jing Guo hand region is counted, and determines the upper the largest number of regions of contiguous pixels of circle
For wrist area, and reject the profile point that the wrist area includes；
(2.6) remaining region outer to circle find local maxima mass center away from profile point, and by the profile point labeled as referring to
Point；
(2.7) it will test the number N of finger tip1Compared with the finger tip number N being had detected that in step (2.4), if the two
It is equal, then it realizes the detection of finger tip, obtains the number and location of finger tip；Otherwise it will test the number N of finger tip1It is updated to step
(2.4) the finger tip number N having detected that in, and return step (2.4), continue the detection of finger tip.
It is described that hand area is calculated according to the coordinate of the area of hand region and each homalographic block in step (2.2)
The mass center in domain refers to: the mass center (x of hand region is calculated by formula (1)g,yg):
Wherein, S is the area of hand region, xi、yiIt is i-th of homalographic block respectively in X-direction and Y direction coordinate.
In step (2.3), the distance of the profile point for calculating all samplings to mass center, and calculate average mass center away from
Refer to: calculating the profile point (x of all samplingspoint_i,ypoint_i) arriving the Euclidean distance of mass center, average mass center with formula (2) away from being counted
It calculates:
Wherein, (xg,yg) be hand region mass center.
In step (2.4), the finger tip number N that the basis has detected that determines extended centroid away from DextRefer to: it is described
The finger tip number N and extended centroid detected is away from DextThe condition of satisfaction are as follows:
0≤N<3,Dext=1.5 × Davg
3≤N≤5,Dext=1.2 × Davg
If finger tip number N > 5 having detected that, return to the first step；Wherein, DavgFor average mass center away from.
In step (2.6), the outer remaining region of described pair of circle find local maxima mass center away from profile point, and by the wheel
Exterior feature point refers to labeled as finger tip: by traversing finger tip profile point, when the mass center of 10 profile points continuous in profile point is away from gradually increasing
Added-time records current maximum, when mass center of the mass center away from continuous 10 profile points is away from starting to reduce, will currently record most
Big value is as local maxima mass center away from local maxima mass center is then labeled as finger tip away from corresponding profile point.
Compared with prior art, the invention has the advantages that with the utility model has the advantages that
1, it is different from existing other methods, this method is not needed using special camera or special installation, not when implementing yet
It needs to carry out special marking to manpower, manpower can freely be moved before camera.
Wherein, when carrying out hand region segmentation, the light stream region of variation of the video scene of capture is visualized, and passes through
The mode of YCbCr Complexion filter device and the screening of profile descending is constructed, effectively realizes the hand region segmentation under complex environment, it is maximum
Degree avoids the interference of ambient noise, can cope in scene the case where including a large amount of class area of skin color.
When carrying out finger tip detection to hand region, this method constructs hand region homalographic block models, passes through the mould
The parser of type calculates average mass center away from (Davg) and extended centroid away from (Dext), and with DavgWith different DextIt is drawn for radius
Circle carries out wrist direction selection and profile point screening, eventually by local maxima mass center is calculated away from fingertip location is found, effectively arranges
Except erroneous detection measuring point in circle, to effectively reject the interference of the noise informations to finger tip detection such as wrist, the tiny protrusion of hand, it is ensured that refer to
Cusp is located at outside circle, realizes the finger tip detection of robust.Further, since circle has rotational invariance, therefore when palm is rotated into not
When with angle, this method can still correctly detect finger tip.
2, the Fingertip Detection not only can effectively realize the segmentation that hand region is carried out under complex environment, but also the inspection
Survey method strong robustness, can also correctly detect finger tip when manpower is freely moved before camera under complex environment, to mention
The accuracy and validity of high finger tip detection.
Detailed description of the invention
Fig. 1 is the flow diagram of Fingertip Detection of the present invention；
Fig. 2 is the method flow diagram that the first step carries out hand region segmentation from complex environment；
Fig. 3 is the method flow diagram that second step carries out finger tip detection to the hand region split；
Fig. 4 is hand region display model figure when the palm the five fingers open；
Fig. 5 is hand region display model figure when fist is presented in palm；
Fig. 6 is the schematic diagram that hand region is reduced to geometry appearance model when fist is presented in palm；
Hand region is reduced to the schematic diagram of geometry appearance model when Fig. 7 is palm list finger tip；
Specific embodiment
The present invention is described in further detail with specific embodiment with reference to the accompanying drawing.
Embodiment
As shown in Figures 1 to 3, Fingertip Detection under complex environment of the present invention, is detected for the finger tip to hand；Its
It is characterized in that: including following two step:
The first step, capturing scenes information, by calculating dense optical flow information corresponding with scene information and carrying out binaryzation
Pretreatment obtains the scene image containing hand region；Construct Complexion filter device again, to the scene image containing hand region into
The segmentation of row hand region, obtains hand region；
Second step constructs the model of each posture hand region using homalographic block, according to the areal calculation hand of hand region
The mass center in portion region；Hand region contour is sampled again, all configuration sampling points is calculated to the distance of mass center, and calculates
Average mass center away from；Then determined extended centroid away from expanding and using mass center as the center of circle according to the last round of finger tip number having detected that
Exhibition mass center is round away from picture is carried out for radius；The profile point and the upper the largest number of wrist areas of contiguous pixels of circle in circle are finally rejected,
And circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；This wheel is detected to the number of finger tip
Compared with the last round of finger tip number having detected that, if the two is equal, realizes the detection of finger tip, otherwise detect this wheel
The number of finger tip is updated to the finger tip number having detected that, come determine extended centroid away from and carry out picture circle, continue the detection of finger tip.
In the first step, the capturing scenes information is gone forward side by side by calculating dense optical flow information corresponding with scene information
Row binaryzation pre-processes to obtain the scene image containing hand region；Complexion filter device is constructed again, to the field containing hand region
Scape image carries out the segmentation of hand region, obtains hand region and refers to: the following steps are included:
(1.1) capturing scenes information, and calculate corresponding dense optical flow information；Although the calculation amount one of dense optical flow
As it is all very big, but in the method, be intended merely to obtain hand candidate approximate region since the present invention calculates dense optical flow,
We carry out optical flow computation using double-deck pyramid, and set biggish search window (15x15).
(1.2) Optic flow information is traversed, maximum light stream value is found, and utilize the maximum light stream value, in X-axis and Y direction pair
All light streams carry out Regularization.
(1.3) according to the light stream after regularization, tone and saturation degree along the changed region in light stream direction are calculated, and
Mark different color values.
(1.4) given threshold converts bianry image for the light stream region of variation of color mark according to threshold value, in conjunction with logic
Operation and mathematical morphological operation obtain the scene image containing hand region.
(1.5) according to human body complexion Clustering features, YCbCr Complexion filter device is constructed, rejects redundancy color and luminance information；
Under normal circumstances, candidate hand region image includes RGB color information, can convert it to YCbCr color according to formula (3)
Information.
It, can will be public in conjunction with a large amount of experimental result simultaneously as the distribution of human body complexion has apparent Clustering features
The long and narrow colour band that formula (4) provides constructs YCbCr Complexion filter device as complexion model.
(1.6) all profiles of the scene image containing hand region are passed after being filtered to YCbCr Complexion filter device
Emission reduction sequence finds largest connected region as hand region, realizes and carry out hand region to the scene image containing hand region
Segmentation.
In second step, the model that each posture hand region is constructed using homalographic block, to determine hand region
Area, to calculate the mass center of hand region；Hand region contour is sampled again, calculates all configuration sampling points to mass center
Distance, and calculate average mass center away from；Then determined according to the last round of finger tip number having detected that extended centroid away from, and with
Mass center is the center of circle, and extended centroid is round away from picture is carried out for radius；The upper contiguous pixels number of profile point and circle in finally rejecting is round is most
More wrist areas, and circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；This wheel is detected
To finger tip number compared with the last round of finger tip number having detected that, if the two is equal, realize the detection of finger tip, otherwise will
This wheel detects the finger tip number that the number of finger tip is updated to have detected that, come determine extended centroid away from and carry out picture circle, after
The detection of continuous finger tip refers to: the following steps are included:
(2.1) homalographic is obtained according to the different postures of hand region with homalographic block building hand region display model
The quantity of block, to determine the area of hand region；As shown in figure 4, when the palm the five fingers open, in conjunction with hand Degrees of Freedom Model,
Hand region display model can be indicated with 26 homalographic blocks；As shown in figure 5, then can be used 16 when fist is presented in palm
A homalographic block indicates its display model.For ease of calculation, by taking fist display model as an example, which can be further simplified
For geometry appearance model, as shown in Figure 6.
(2.2) mass center of hand region is calculated according to the coordinate of the area of hand region and each homalographic block；Specifically
It is the mass center (x that hand region is calculated by formula (1)g,yg):
Wherein, S is the area of hand region, xi、yiIt is i-th of homalographic block respectively in X-direction and Y direction coordinate.
(2.3) intensive sampling is carried out to hand region contour, calculates the profile point of all samplings to the distance of mass center, and count
Calculate average mass center away from；Calculate the profile point (x of all samplingspoint_i,ypoint_i) arrive the Euclidean distance of mass center, average mass center away from
It is calculated with formula (2):
Wherein, (xg,yg) be hand region mass center.
(2.4) determine extended centroid away from D according to the finger tip number N having detected thatext, and using mass center as the center of circle, extended centroid
Away from DextPicture circle is carried out for radius；And rejecting includes the profile point in the circle.Wherein, the initial value of N is set as 0；；It has detected
The finger tip number N and extended centroid arrived is away from DextThe condition of satisfaction are as follows:
0≤N<3,Dext=1.5 × Davg
3≤N≤5,Dext=1.2 × Davg
If finger tip number N > 5 having detected that, return to the first step；Wherein, DavgFor average mass center away from.
Based on hand display model, when different finger numbers are presented in palm, the number of homalographic block is different, therefore with
Different extended centroids are justified away from drawing for radius, effectively by wrist area and can reject the interference of tiny protruding point.Outside in conjunction with hand
Geometrical model and formula (1), formula (2) are seen, when finger number is less than 3, extended centroid is away from the average matter that should be 1.5 times
The heart is away from (Dext=1.5 × Davg)；When finger number is greater than 3 but when less than 5, extended centroid is away from being 1.2 times of mass centers that are averaged away from (Dext
=1.2 × Davg)；When finger number is greater than 5, terminates finger tip detection, do not do with extended centroid away from the circle for radius.With Fig. 6 institute
For the fist geometry appearance model shown, mass center is located at center, is justified using 1.5 times of average mass centers away from drawing as radius, the circle
By comprising homalographic block all on fist, so that profile point all on fist all be rejected, go out finger tip without erroneous detection.Again
Single finger tip geometry appearance model as shown in Figure 7 is justified using 1.5 times of average mass centers away from drawing as radius, in addition to the finger tip of protrusion,
His most of area block will be all included in extended centroid away from the circle for radius, to effectively reject the interference of tiny protrusion
Point, by it is subsequent calculate determine with extended centroid away from the circle for radius outside fingertip location.
(2.5) number of pixels of the circle Jing Guo hand region is counted, and determines the upper the largest number of regions of contiguous pixels of circle
For wrist area, the direction rejects the profile point that the wrist area includes without finger tip detection.
(2.6) remaining region outer to circle find local maxima mass center away from profile point, and by the profile point labeled as referring to
Point.Step (2.6) specifically be such that by traverse finger tip profile point, when 10 profile points continuous in profile point mass center away from
When gradually increasing, current maximum is recorded, when mass center of the mass center away from continuous 10 profile points is away from starting to reduce, will currently have been remembered
The maximum value of record is as local maxima mass center away from local maxima mass center is then labeled as finger tip away from corresponding profile point.
(2.7) it will test the number N of finger tip1Compared with the finger tip number N being had detected that in step (2.4), if the two
It is equal, then it realizes the detection of finger tip, obtains the number and location of finger tip；Otherwise it will test the number N of finger tip1It is updated to step
(2.4) the finger tip number N having detected that in, and return step (2.4), continue the detection of finger tip.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (2)
1. Fingertip Detection under a kind of complex environment is detected for the finger tip to hand；It is characterized by comprising following
Two steps:
The first step, capturing scenes information are located in advance by calculating dense optical flow information corresponding with scene information and carrying out binaryzation
Reason obtains the scene image containing hand region；Complexion filter device is constructed again, and hand is carried out to the scene image containing hand region
The segmentation in portion region, obtains hand region；
Second step constructs the model of each posture hand region using homalographic block, according to the areal calculation hand area of hand region
The mass center in domain；Hand region contour is sampled again, all configuration sampling points is calculated to the distance of mass center, and calculates average
Mass center away from；Then determined extended centroid away from extending matter and using mass center as the center of circle according to the last round of finger tip number having detected that
The heart is round away from picture is carried out for radius；The finally profile point in rejecting circle and the upper the largest number of wrist areas of contiguous pixels of circle, and
Circle is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；By this wheel detect the number of finger tip with it is upper
The finger tip number that one wheel has detected that compares, if the two is equal, realizes the detection of finger tip, this wheel is otherwise detected finger tip
Number be updated to the finger tip number having detected that, come determine extended centroid away from and carry out picture circle, continue the detection of finger tip；
In the first step, the capturing scenes information, by calculating dense optical flow information corresponding with scene information and carrying out two
Value pre-processes to obtain the scene image containing hand region；Complexion filter device is constructed again, to the scene figure containing hand region
Segmentation as carrying out hand region, obtains hand region and refers to: the following steps are included:
(1.1) capturing scenes information, and calculate corresponding dense optical flow information；
(1.2) Optic flow information is traversed, finds maximum light stream value, and utilize the maximum light stream value, in X-axis and Y direction to all
Light stream carries out Regularization；
(1.3) according to the light stream after regularization, tone and saturation degree along the changed region in light stream direction are calculated, and is marked
Different color values；
(1.4) given threshold converts bianry image for the light stream region of variation of color mark according to threshold value, in conjunction with logical operation
And mathematical morphological operation, obtain the scene image containing hand region；
(1.5) according to human body complexion Clustering features, YCbCr Complexion filter device is constructed, rejects redundancy color and luminance information；
(1.6) row that successively decreases is carried out to all profiles of the scene image containing hand region after filtering to YCbCr Complexion filter device
Sequence finds largest connected region as hand region, realizes point for carrying out hand region to the scene image containing hand region
It cuts；
In second step, the model that each posture hand region is constructed using homalographic block, to determine the area of hand region,
To calculate the mass center of hand region；Hand region contour is sampled again, calculate all configuration sampling points to mass center distance,
And calculate average mass center away from；Then determined extended centroid away from and with mass center according to the last round of finger tip number having detected that
For the center of circle, extended centroid is away from carrying out picture circle for radius；Finally the profile point in rejecting circle and the upper contiguous pixels of circle are the largest number of
Wrist area, and circle it is outer find local maxima mass center away from profile point, then the point is labeled as finger tip；This wheel is detected into finger
The number of point if the two is equal, realizes the detection of finger tip, otherwise by this compared with the last round of finger tip number having detected that
Wheel detect the finger tip number that the number of finger tip is updated to have detected that, come determine extended centroid away from and carry out picture circle, continue to refer to
The detection of point refers to: the following steps are included:
(2.1) homalographic block is obtained according to the different postures of hand region with homalographic block building hand region display model
Quantity, to determine the area of hand region；
(2.2) mass center of hand region is calculated according to the coordinate of the area of hand region and each homalographic block；
(2.3) intensive sampling is carried out to hand region contour, calculates the profile point of all samplings to the distance of mass center, and calculate
Average mass center away from；
(2.4) determine extended centroid away from D according to the finger tip number N having detected thatext, and using mass center as the center of circle, extended centroid away from
DextPicture circle is carried out for radius；And rejecting includes the profile point in the circle；Wherein, the initial value of N is set as 0；
(2.5) number of pixels of the circle Jing Guo hand region is counted, and determines that the upper the largest number of regions of contiguous pixels of circle are hand
Wrist region, and reject the profile point that the wrist area includes；
(2.6) remaining region outer to circle find local maxima mass center away from profile point, and the profile point is labeled as finger tip；
(2.7) it will test the number N of finger tip1Compared with the finger tip number N being had detected that in step (2.4), if the two is equal,
The detection for then realizing finger tip obtains the number and location of finger tip；Otherwise it will test the number N of finger tip1It is updated to step (2.4)
In the finger tip number N that has detected that, and return step (2.4) continues the detection of finger tip；
It is described that hand region is calculated according to the coordinate of the area of hand region and each homalographic block in step (2.2)
Mass center refers to: the mass center (x of hand region is calculated by formula (1)g,yg):
Wherein, S is the area of hand region, xi、yiIt is i-th of homalographic block respectively in X-direction and Y direction coordinate；N is
The sum of hand region profile point；
In step (2.3), the distance of the profile point for calculating all samplings to mass center, and average mass center is calculated away from being
Refer to: calculating the profile point (x of all samplingspoint_i,ypoint_i) arriving the Euclidean distance of mass center, average mass center with formula (2) away from being counted
It calculates:
Wherein, (xg,yg) be hand region mass center；
In step (2.4), the finger tip number N that the basis has detected that determines extended centroid away from DextRefer to: described to have detected
The finger tip number N and extended centroid arrived is away from DextThe condition of satisfaction are as follows:
0≤N<3,Dext=1.5 × Davg
3≤N≤5,Dext=1.2 × Davg
If finger tip number N > 5 having detected that, return to the first step；Wherein, DavgFor average mass center away from.
2. Fingertip Detection under complex environment according to claim 1, it is characterised in that: in step (2.6), in step
Suddenly in (2.6), the outer remaining region of described pair of circle find local maxima mass center away from profile point, and by the profile point labeled as referring to
Point refers to: by traversing finger tip profile point, when the mass center of 10 profile points continuous in profile point is away from gradually increasing, record is current
Maximum value, when mass center of the mass center away from continuous 10 profile points is away from starting to reduce, using the maximum value currently recorded as part
Maximum mass center is away from local maxima mass center is then labeled as finger tip away from corresponding profile point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510700440.2A CN105335711B (en) | 2015-10-22 | 2015-10-22 | Fingertip Detection under a kind of complex environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510700440.2A CN105335711B (en) | 2015-10-22 | 2015-10-22 | Fingertip Detection under a kind of complex environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105335711A CN105335711A (en) | 2016-02-17 |
| CN105335711B true CN105335711B (en) | 2019-01-15 |
Family
ID=55286227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510700440.2A Active CN105335711B (en) | 2015-10-22 | 2015-10-22 | Fingertip Detection under a kind of complex environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105335711B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR3048108B1 (en) * | 2016-02-24 | 2019-11-01 | Hins | METHOD FOR RECOGNIZING THE DISPOSITION OF A HAND IN AN IMAGE STREAM |
| CN106408579B (en) * | 2016-10-25 | 2019-01-29 | 华南理工大学 | A kind of kneading finger tip tracking based on video |
| CN106529432B (en) * | 2016-11-01 | 2019-05-07 | 山东大学 | A kind of hand region dividing method of depth integration conspicuousness detection and priori knowledge |
| CN107180224B (en) * | 2017-04-10 | 2020-06-19 | 华南理工大学 | Finger motion detection and positioning method based on space-time filtering and joint space Kmeans |
| CN109118529B (en) * | 2018-08-13 | 2022-06-03 | 四川长虹电器股份有限公司 | Screw hole image rapid positioning method based on vision |
| CN112114675B (en) * | 2020-09-29 | 2023-05-26 | 陕西科技大学 | Gesture control-based non-contact elevator keyboard using method |
| CN112906563A (en) * | 2021-02-19 | 2021-06-04 | 山东英信计算机技术有限公司 | Dynamic gesture recognition method, device and system and readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103034333A (en) * | 2012-12-18 | 2013-04-10 | 福建华映显示科技有限公司 | Gesture recognition device and gesture recognition method |
| CN103488972A (en) * | 2013-09-09 | 2014-01-01 | 西安交通大学 | Method for detection fingertips based on depth information |
| US20140161309A1 (en) * | 2012-12-07 | 2014-06-12 | Chunghwa Picture Tubes, Ltd. | Gesture recognizing device and method for recognizing a gesture |
| CN104899600A (en) * | 2015-05-28 | 2015-09-09 | 北京工业大学 | Depth map based hand feature point detection method |
-
2015
- 2015-10-22 CN CN201510700440.2A patent/CN105335711B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140161309A1 (en) * | 2012-12-07 | 2014-06-12 | Chunghwa Picture Tubes, Ltd. | Gesture recognizing device and method for recognizing a gesture |
| CN103034333A (en) * | 2012-12-18 | 2013-04-10 | 福建华映显示科技有限公司 | Gesture recognition device and gesture recognition method |
| CN103488972A (en) * | 2013-09-09 | 2014-01-01 | 西安交通大学 | Method for detection fingertips based on depth information |
| CN104899600A (en) * | 2015-05-28 | 2015-09-09 | 北京工业大学 | Depth map based hand feature point detection method |
Non-Patent Citations (1)
| Title |
|---|
| 基于单目视觉的手指识别跟踪系统设计;南睿;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140315(第03期);I138-896 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105335711A (en) | 2016-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105335711B (en) | Fingertip Detection under a kind of complex environment | |
| CN105718878B (en) | The aerial hand-written and aerial exchange method in the first visual angle based on concatenated convolutional neural network | |
| Mukherjee et al. | Fingertip detection and tracking for recognition of air-writing in videos | |
| Zhou et al. | A novel finger and hand pose estimation technique for real-time hand gesture recognition | |
| CN108052896B (en) | Human body behavior identification method based on convolutional neural network and support vector machine | |
| JP6079832B2 (en) | Human computer interaction system, hand-to-hand pointing point positioning method, and finger gesture determination method | |
| CN105739702B (en) | Multi-pose finger tip tracking for natural human-computer interaction | |
| Liu et al. | Hand posture recognition using finger geometric feature | |
| CN109255324A (en) | Gesture processing method, interaction control method and equipment | |
| CN101561710B (en) | Man-machine interaction method based on estimation of human face posture | |
| US20190034714A1 (en) | System and method for detecting hand gestures in a 3d space | |
| CN109948542A (en) | Gesture identification method, device, electronic equipment and storage medium | |
| Agrawal et al. | A survey on manual and non-manual sign language recognition for isolated and continuous sign | |
| Feng et al. | Real-time fingertip tracking and detection using Kinect depth sensor for a new writing-in-the air system | |
| CN106200971A (en) | Man-machine interactive system device based on gesture identification and operational approach | |
| Chen et al. | A novel illumination-robust hand gesture recognition system with event-based neuromorphic vision sensor | |
| Yao et al. | Real-time hand pose estimation from RGB-D sensor | |
| CN103995595A (en) | Game somatosensory control method based on hand gestures | |
| CN110458235B (en) | Motion posture similarity comparison method in video | |
| Xu et al. | Robust hand gesture recognition based on RGB-D Data for natural human–computer interaction | |
| CN105261038B (en) | Finger tip tracking based on two-way light stream and perception Hash | |
| CN110956099A (en) | Dynamic gesture instruction identification method | |
| Ranawat et al. | Hand gesture recognition based virtual mouse events | |
| CN102073878B (en) | Non-wearable finger pointing gesture visual identification method | |
| Simion et al. | Finger detection based on hand contour and colour information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |