CN102867172A - Human eye positioning method, system and electronic equipment - Google Patents
Human eye positioning method, system and electronic equipment Download PDFInfo
- Publication number
- CN102867172A CN102867172A CN2012103088763A CN201210308876A CN102867172A CN 102867172 A CN102867172 A CN 102867172A CN 2012103088763 A CN2012103088763 A CN 2012103088763A CN 201210308876 A CN201210308876 A CN 201210308876A CN 102867172 A CN102867172 A CN 102867172A
- Authority
- CN
- China
- Prior art keywords
- sample image
- human
- cascade classifier
- eye
- rectangular area
- 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.)
- Granted
Links
Images
Abstract
The invention belongs to the technical fields of digital image processing and pattern recognition, and provides a human eye positioning method, a human eye positioning system and electronic equipment. The method comprises the following steps of: extracting a face area from an image currently to be recognized by utilizing a first cascade classifier for face detection, and selecting a corresponding second cascade classifier from two or more than two second cascade classifier for human eye detection at different distances according to the size of the face area; and positioning human eyes in the image currently to be recognized by utilizing the selected second cascade classifier. The characteristics of the human eyes at different distances away from image acquisition equipment are different, so that the method, the system and the electronic equipment can be well adapted to the human eye detection positioning tasks of different distances by utilizing the cascade classifiers suitable for different distances, the human eye positioning accuracy is improved, and the phenomena of false detection and missing detection are avoided.
Description
Technical field
The invention belongs to Digital Image Processing and mode identification technology, relate in particular to a kind of human-eye positioning method, system and electronic equipment.
Background technology
In recent years, human eye detection and location technology are widely used in the aspects such as medical application, video conference, police criminal detection as the important technology in computer vision and the area of pattern recognition.
Prior art provides a kind of human-eye positioning method of intensity-based sciagraphy, the method is the human face region in the detected image at first, afterwards on human face region, people's face gray level image is carried out projection on level and the vertical direction, find respectively the specific change point of Gray Projection on this both direction, and in conjunction with priori the position of specific change point is comprehensively judged, thereby obtain the position of human eye.But the method is extremely responsive to the noise that exists in the variation of human face posture and the image, detect and locating effect relatively poor, and only just produce effect during the close together between video camera and people's face.
For this reason, prior art provides another kind of human-eye positioning method based on class Ha Er (Haar) feature and self-adaptation enhancing (Adaboost) algorithm.The method is levied a large amount of human eye samples and the training of non-human eye sample based on the class Lis Hartel, utilize self-adaptive enhancement algorithm to extract some Weak Classifiers of better performances, and the Weak Classifier that extracts carried out cascade, form final strong classifier, and utilize this strong classifier to treat human eye in the recognition image and detect and locate.Human-eye positioning method with respect to the intensity-based sciagraphy, the detection of this kind method is effective, locating speed is fast, but still the distance limit between unresolved video camera and the people's face, when distant between video camera and the people's face, because the human eye feature that detects is fuzzyyer, situation undetected, flase drop occurs easily.
Disclosed above-mentioned information is only for increasing the understanding to background technology of the present invention in this background technology this part, so it may comprise the prior art known to persons of ordinary skill in the art that does not consist of this state.
Summary of the invention
The purpose of the embodiment of the invention is to provide a kind of human-eye positioning method, be intended to solve prior art when the self-adaptive enhancement algorithm of seeking peace based on the class Lis Hartel carries out the human eye location, if distant between video camera and the people's face, because human eye feature is fuzzyyer, is easy to occur problem undetected, flase drop.
The embodiment of the invention is achieved in that a kind of human-eye positioning method, said method comprising the steps of:
Utilize the first cascade classifier that is used for the detection of people's face to extract human face region from current image to be identified, according to the size of described human face region, select corresponding described the second cascade classifier from the second cascade classifier two or more, that be used for the different distance human eye detection;
Described the second cascade classifier that utilization is selected is located the position of human eye in the described current image to be identified.
Another purpose of the embodiment of the invention is to provide a kind of human eye positioning system, and described system comprises:
The second cascade classifier selected cell, be used for utilizing the first cascade classifier that is used for the detection of people's face to extract human face region from current image to be identified, according to the size of described human face region, select corresponding the second cascade classifier from the second cascade classifier two or more, that be used for the different distance human eye detection;
Positioning unit, described the second cascade classifier that is used for utilizing described the second cascade classifier selected cell to select is located the position of human eye of described current image to be identified.
Another purpose of the embodiment of the invention is to provide a kind of electronic equipment, comprise a display unit, described electronic equipment also comprises as mentioned above a human eye positioning system, the described position of human eye of the described current image to be identified that described display unit is used for showing that described human eye positioning system location obtains.
The human-eye positioning method that the embodiment of the invention provides and system utilize the first cascade classifier to extract human face region on the image to be identified, and according to the size of this human face region, select the second suitable cascade classifier to carry out the human eye location.Because the feature of the human eye different from the image capture device distance there are differences, thereby use the human eye detection location tasks that human-eye positioning method that the embodiment of the invention provides and system can better adapt to different distance, improve the accuracy rate of human eye location, avoid the generation of flase drop, the phenomenon such as undetected.
Description of drawings
Fig. 1 is the process flow diagram of the human-eye positioning method that provides of the embodiment of the invention;
Fig. 2 is the exemplary plot of people's face rectangular area of extracting in the human-eye positioning method that provides of the embodiment of the invention;
Fig. 3 is the exemplary plot of the right eye rectangular area that extracts in the human-eye positioning method that provides of the embodiment of the invention;
Fig. 4 is the exemplary plot of the left eye rectangular area that extracts in the human-eye positioning method that provides of the embodiment of the invention;
Fig. 5 is the structure principle chart of the human eye positioning system that provides of the embodiment of the invention;
Fig. 6 is the structural drawing of training unit among Fig. 5;
Fig. 7 is the structural drawing of the second cascade classifier selected cell among Fig. 5;
Fig. 8 is the structural drawing of positioning unit among Fig. 5.
Embodiment
In order to make purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, is not intended to limit the present invention.
Problem for the location existence of existing human eye, the human-eye positioning method that the embodiment of the invention provides is based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace and trains the first cascade classifier and a plurality of the second cascade classifier, utilize the first cascade classifier to extract human face region on the image to be identified, and select the second suitable cascade classifier to carry out the location of human eye according to this human face region.
Fig. 1 shows the flow process of the human-eye positioning method that the embodiment of the invention provides.
In step S11, capturing sample image, and based on the class Lis Hartel self-adaptive enhancement algorithm training sample image of seeking peace, obtain the first cascade classifier of detecting for people's face, and two or more, be used for the second cascade classifier of different distance human eye detection.
Further, if based on the class Lis Hartel self-adaptive enhancement algorithm training sample image of seeking peace, obtain one and be used for closely the second cascade classifier and second cascade classifier that is used for remote human eye detection of human eye detection, then step S11 can further may further comprise the steps:
Step S111: gather people's face sample image, based on the class Lis Hartel self-adaptive enhancement algorithm training of human face sample image of seeking peace, obtain the first cascade classifier that detects for people's face.
Step S112: gather closely left eye sample image and closely right eye sample image, the closely left eye sample image that will gather afterwards carries out flip horizontal, consist of closely human eye sample image storehouse by the closely left eye sample image after the upset and the closely right eye sample image of collection, the closely right eye sample image that perhaps will gather carries out flip horizontal, consists of closely human eye sample image storehouse by the closely right eye sample image that obtains after the upset and the closely left eye sample image of collection.
Step S113: train closely human eye sample image storehouse based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain for the second cascade classifier of human eye detection closely.
In order to improve the accuracy of human eye detection and location, in the embodiment of the invention, before step S113, also need in the human eye sample image storehouse closely, position of human eye alignment in the closely left eye sample image after the upset, position of human eye alignment in the closely right eye sample image that will gather simultaneously, perhaps will be closely in the human eye sample image storehouse, position of human eye alignment in the closely right eye sample image after the upset, position of human eye alignment in the closely left eye sample image that will gather simultaneously, and the closely right eye sample image after the closely left eye sample image that will gather and the upset, or the closely left eye sample image after the closely right eye sample image that gathers and the upset zooms to unified size.
Step S114: gather remote left eye sample image and remote right eye sample image, the remote left eye sample image that will gather afterwards carries out flip horizontal, by the upset after remote left eye sample image and the remote right eye sample image of collection consist of remote human eye sample image storehouse, the remote right eye sample image that perhaps will gather carries out flip horizontal, consists of remote human eye sample image storehouse by the remote right eye sample image after the upset and the remote left eye sample image of collection.
Because when people's range image collecting device is far away, human eye area in the eye image is closely not clear as the eye image and feature is obvious for another example, but obscure mutually with other facial characteristics (as: eyebrow etc.) around the human eye area, if then be still to be used for the closely human eye of the second cascade classifier location remote image of human eye detection, flase drop and the phenomenon such as undetected very easily appear, therefore, in the embodiment of the invention, the closely left eye sample image that gathers includes closely people's left eye region on the face, the closely right eye sample image that gathers includes closely people's right eye region on the face, and the remote left eye sample image that gathers includes remote people left eyebrow zone and left eye region on the face, and the remote right eye sample image of collection includes remote people right eyebrow zone and right eye region on the face.Like this in follow-up human eye location, eyebrow and human eye are detected as an integral body and locate, can reduce false drop rate.
Step S115: train remote human eye sample image storehouse based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain the second cascade classifier for remote human eye detection.
Similarly, in order to improve the accuracy of human eye detection and location, in the embodiment of the invention, before step S 115, also need in the remote human eye sample image storehouse, position of human eye alignment in the remote left eye sample image after the upset, position of human eye alignment in the remote right eye sample image that will gather simultaneously, perhaps with in the remote human eye sample image storehouse, position of human eye alignment in the remote right eye sample image after the upset, position of human eye alignment in the remote left eye sample image that will gather simultaneously, and the remote right eye sample image after the remote left eye sample image that will gather and the upset, or the remote left eye sample image after the remote right eye sample image upset that gathers zooms to unified size.
In step S12, utilize the first cascade classifier that is used for the detection of people's face to extract human face region from current image to be identified, according to the size of human face region, select corresponding the second cascade classifier from the second cascade classifier two or more, that be used for the different distance human eye detection.
In the embodiment of the invention, be used for closely the second cascade classifier and second cascade classifier that is used for remote human eye detection of human eye detection if obtain one in step S11, then step S12 can further may further comprise the steps:
Step S121: utilize the first cascade classifier from current image to be identified, to extract people's face rectangular area, and obtain height, width and the people's face rectangular area positional information in current image to be identified of people's face rectangular area.
Step S122: when the width of people's face rectangular area during more than or equal to a predetermined width value, think that then people's face in the current image to be identified and the distance between the image capture device are in short range, thereby selection is used for closely the second cascade classifier of human eye detection, when the width of people's face rectangular area during less than this predetermined width value, then think people's face in the current image to be identified and the distance between the image capture device in far range, thereby select to be used for the second cascade classifier of remote human eye detection.
Further, if current image to be identified is video frame images, then because the continuity of video frame images, the width of people's face rectangular area is near the predetermined width value time in the continuous multiple frames image, might appear at the phenomenon of beating about the predetermined width value, switch and cause for the second cascade classifier of human eye detection closely and the frequent selection that is used for the second cascade classifier of remote human eye detection.At this moment, beat in order to reduce to switch, in the embodiment of the invention, when being preset with a predetermined width value, also preset a transition value, if current image to be identified is first video frame images, then to the selection of the second cascade classifier as described in the step S122, if current image to be identified is not first video frame images, then step S122 can be again: when the width of people's face rectangular area during greater than predetermined width value and transition value sum, select to be used for closely the second cascade classifier of human eye detection; When the width of people's face rectangular area during less than the difference of predetermined width value and transition value, select to be used for the second cascade classifier of remote human eye detection; Width when people's face rectangular area is less than or equal to predetermined width value and transition value sum, and during more than or equal to the difference of predetermined width value and transition value, select selected the second cascade classifier of last video frame images, that is to say, for the second cascade classifier of human eye detection closely if last video frame images is selected, then the current video two field picture is also selected for the second cascade classifier of human eye detection closely, if be that then the current video two field picture is also selected the second cascade classifier for remote human eye detection for the second cascade classifier of remote human eye detection and last video frame images is selected.
In step S13, the second cascade classifier that utilization is selected is located the position of human eye in the current image to be identified.Further, at first to locate the right eye position, to relocate the left eye position as example, step S13 can may further comprise the steps:
Step S131: utilize the second cascade classifier of selecting, in the first half of the human face region that step S12 extracts, extract the right eye rectangular area, obtain size and the positional information in human face region thereof of right eye rectangular area, and the coordinate of center on facial image to be identified that obtains the right eye rectangular area according to size and the positional information in human face region thereof of right eye rectangular area, this centre coordinate is the right eye position in the image current to be identified that step S13 location obtains.
Step S132: the first half of the human face region that step S12 is extracted is carried out flip horizontal, utilize the second cascade classifier of selecting, extract the left eye rectangular area in the human face region after upset, obtain the size of left eye rectangular area and the positional information in the human face region after upset thereof, and according to the size of left eye rectangular area and the positional information in the human face region after upset thereof, obtain the coordinate of center on unturned facial image to be identified of left eye rectangular area, this centre coordinate is the left eye position in the image current to be identified that step S13 location obtains.
Because in some cases, utilize the second cascade classifier of selecting on upset or unturned human face region, to detect respectively left eye and right eye, that is to say, might in upset and unturned human face region, navigate to two left eye positions or two right eye positions, then at this moment, behind execution in step S132, if in the left-half of the first half of human face region, navigate to two right eye positions, the right eye position that then navigates to so that step S131 does not overturn human face region is final right eye position, if navigate to two left eye positions in the right half part of the first half of human face region, the left eye position that then navigates to behind the step S132 upset human face region is as final left eye position.
Certainly, if at first locate the left eye position, relocate the right eye position, then the detailed step of step S13 is similar with step S132 to step S131, for avoiding repetition, does not repeat them here.Simultaneously, when practical application, also can prestore the first cascade classifier and two or more, be used for the second cascade classifier of different distance human eye detection, thereby need not execution in step S11 and directly execution in step S12 and step S13.
The human-eye positioning method that the embodiment of the invention provides is based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace and trains the first cascade classifier and a plurality of the second cascade classifier, utilize the first cascade classifier to extract human face region on the image to be identified, and select the second suitable cascade classifier to carry out the human eye location according to this human face region.Because the feature of the human eye different from the image capture device distance there are differences, thereby the human-eye positioning method of using the embodiment of the invention and providing can better adapt to the human eye detection location tasks of different distance, improve the accuracy rate of human eye location, avoid the generation of flase drop, the phenomenon such as undetected.
For instance, suppose one section continuous people's face video is carried out the human eye location, through people's face rectangular area that step S121 extracts as shown in Figure 2, the height of this people's face rectangular area is that h, width are w, the coordinate of this people's face rectangular area left upper apex D in the XOY coordinate system is that (x, y) namely is the positional information of people's face rectangular area in current image to be identified.Suppose that the predetermined width value is T, transition value is △ t, then in step S122, if w＞T+ △ t, select to be used for closely the second cascade classifier of human eye detection, if w＜T-△ t selects the second cascade classifier for remote human eye detection, if T-△ is t≤w≤and T+ △ t, select selected the second cascade classifier of former frame image.Afterwards, in step S131, in the first half of people's face rectangular area, extract the right eye rectangular area as shown in Figure 3, the height of this right eye rectangular area is that h1, width are w1, the coordinate of this people's face rectangular area left upper apex in the XOY coordinate system is (x+u1, y+v1), thereby the coordinate of center on facial image to be identified that obtains the right eye rectangular area is (x+u1+w1/2, y+v1+h1/2), this centre coordinate is the right eye position that the location obtains.Afterwards, in step S132, the first half of people's face rectangular area is carried out flip horizontal, utilize the second cascade classifier of selecting, extract the left eye rectangular area as shown in Figure 4 in people's face rectangular area after upset, the height of this left eye rectangular area is h2, width is w2, coordinate in place, the people's face rectangular area XOY coordinate system of this left eye rectangular area left upper apex after upset is (x+u2, y+v2), thereby the coordinate of center on unturned facial image to be identified that obtains the left eye rectangular area is (x+w-u2-w2/2, y+v2+h2/2), this centre coordinate is the left eye position that the location obtains.
Fig. 5 shows the structural principle of the human eye positioning system that the embodiment of the invention provides, and for convenience of explanation, only shows the part relevant with the embodiment of the invention.
The human eye positioning system that the embodiment of the invention provides comprises: training unit 11, be used for capturing sample image, and based on the class Lis Hartel self-adaptive enhancement algorithm training sample image of seeking peace, obtain the first cascade classifier of detecting for people's face, and two or more, be used for the second cascade classifier of different distance human eye detection; The second cascade classifier selected cell 12, the first cascade classifier that is used for utilizing training unit 11 to obtain extracts human face region from current image to be identified, according to size Selection training unit 11 second cascade classifiers that obtain, corresponding of human face region; Positioning unit 13, the second cascade classifier that is used for utilizing the second cascade classifier selected cell 12 to select is located the position of human eye of current image to be identified.
Fig. 6 shows the structure of training unit 11 among Fig. 5.
Particularly, training unit 11 can comprise: the first training module 111, be used for to gather people's face sample image, and based on the class Lis Hartel self-adaptive enhancement algorithm training of human face sample image of seeking peace, obtain the first cascade classifier that detects for people's face; Closely human eye sample image storehouse acquisition module 112, be used for gathering closely left eye sample image and closely right eye sample image, the closely left eye sample image that will gather afterwards carries out flip horizontal, consist of closely human eye sample image storehouse by the closely left eye sample image after the upset and the closely right eye sample image of collection, the closely right eye sample image that perhaps will gather carries out flip horizontal, consists of closely human eye sample image storehouse by the closely right eye sample image after the upset and the closely left eye sample image of collection; The second training module 113 is used for based on the class Lis Hartel closely human eye sample image storehouse that self-adaptive enhancement algorithm trains human eye sample image storehouse acquisition module 112 closely to obtain of seeking peace, and obtains for the second cascade classifier of human eye detection closely; Remote human eye sample image storehouse acquisition module 114, be used for gathering remote left eye sample image and remote right eye sample image, the remote left eye sample image that will gather afterwards carries out flip horizontal, consist of remote human eye sample image storehouse by the remote left eye sample image after the upset and the remote right eye sample image of collection, the remote right eye sample image that perhaps will gather carries out flip horizontal, consists of remote human eye sample image storehouse by the remote right eye sample image after the upset and the remote left eye sample image of collection; The 3rd training module 115, the remote human eye sample image storehouse for the self-adaptive enhancement algorithm of seeking peace based on the class Lis Hartel trains remote human eye sample image storehouse acquisition module 114 to obtain obtains the second cascade classifier for remote human eye detection.
Fig. 7 shows the structure of the second cascade classifier selected cell 12 among Fig. 5.
Particularly, the second cascade classifier selects module 12 to comprise: extraction module 121, the first cascade classifier that is used for utilizing training unit 11 to obtain extracts people's face rectangular area from current image to be identified, and obtains height, width and the people's face rectangular area positional information in current image to be identified of people's face rectangular area; Select module 122, when being used for width when people's face rectangular area that extraction module 121 obtains more than or equal to a predetermined width value, select closely the second cascade classifier of human eye detection that is used for that training unit 11 obtains, when the width of people's face rectangular area during less than this predetermined width value, the second cascade classifier that is used for remote human eye detection of selecting training unit 11 to obtain.And if current image to be identified is video frame images, then select the selection step of module 122 as mentioned above, do not repeat them here.
Fig. 8 shows the structure of positioning unit 13 among Fig. 5.
Particularly, positioning unit 13 can comprise: right eye position locating module 131, be used for utilizing the second cascade classifier of the second cascade classifier selected cell 12 selections, in the first half of the human face region that the second cascade classifier selected cell 12 extracts, extract the right eye rectangular area, obtain size and the positional information in human face region thereof of right eye rectangular area, and the coordinate of center on facial image to be identified that obtains the right eye rectangular area according to size and the positional information in human face region thereof of right eye rectangular area, this centre coordinate is the right eye position in the image current to be identified that positioning unit 13 location obtain; Left eye position locating module 132, the first half for the human face region that the second cascade classifier selected cell 12 is extracted is carried out flip horizontal, the second cascade classifier that utilizes the second cascade classifier selected cell 12 to select, extract the left eye rectangular area in the human face region after upset, obtain the size of left eye rectangular area and the positional information in the human face region after upset thereof, and according to the size of left eye rectangular area and the positional information in the human face region after upset thereof, obtain the coordinate of center on unturned facial image to be identified of left eye rectangular area, this centre coordinate is the left eye position that positioning unit 13 is located in the image current to be identified that obtains.
Certainly, if at first locate the left eye position, relocate the right eye position, then the structure of positioning unit 13 and shown in Figure 5 similar for avoiding repetition, does not repeat them here.Simultaneously, when practical application, the second cascade classifier selected cell 12 also can prestore the first cascade classifier with two or more, for the second cascade classifier of different distance human eye detection, and then the human eye positioning system of this moment can not comprise training unit 11.
The embodiment of the invention also provides a kind of electronic equipment, comprises a display unit and human eye positioning system as mentioned above, and display unit is used for the position of human eye that demonstration human eye positioning system is located the current image to be identified that obtains.
The human-eye positioning method that the embodiment of the invention provides and system utilize the first cascade classifier to extract human face region on the image to be identified, and according to the size of this human face region, select the second suitable cascade classifier to carry out the human eye location.Because the feature of the human eye different from the image capture device distance there are differences, thereby use the human eye detection location tasks that human-eye positioning method that the embodiment of the invention provides and system can better adapt to different distance, improve the accuracy rate of human eye location, avoid the generation of flase drop, the phenomenon such as undetected.
One of ordinary skill in the art will appreciate that all or part of step that realizes in above-described embodiment method is can control relevant hardware by program to finish, described program can be in being stored in a computer read/write memory medium, described storage medium is such as ROM/RAM, disk, CD etc.
The above only is preferred embodiment of the present invention, not in order to limiting the present invention, all any modifications of doing within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.
Claims (14)
1. a human-eye positioning method is characterized in that, said method comprising the steps of:
Utilize the first cascade classifier that is used for the detection of people's face to extract human face region from current image to be identified, according to the size of described human face region, select corresponding described the second cascade classifier from the second cascade classifier two or more, that be used for the different distance human eye detection;
Described the second cascade classifier that utilization is selected is located the position of human eye in the described current image to be identified.
2. human-eye positioning method as claimed in claim 1 is characterized in that, before the first cascade classifier that described utilization detects for people's face extracted the step of human face region from current image to be identified, described method was further comprising the steps of:
Capturing sample image, and train described sample image based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace obtains the first cascade classifier of detecting for people's face, and two or more, be used for the second cascade classifier of different distance human eye detection.
3. human-eye positioning method as claimed in claim 2, it is characterized in that, described capturing sample image, and train described sample image based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain the first cascade classifier of detecting for people's face, and two or more, the step that is used for the second cascade classifier of different distance human eye detection further may further comprise the steps:
Gather people's face sample image, train described people's face sample image based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain the first cascade classifier that detects for people's face;
Gather closely left eye sample image and closely right eye sample image, the described closely left eye sample image that will gather afterwards carries out flip horizontal, consist of closely human eye sample image storehouse by the closely left eye sample image after the upset and the described closely right eye sample image of collection, the described closely right eye sample image that perhaps will gather carries out flip horizontal, consists of closely human eye sample image storehouse by the closely right eye sample image after the upset and the described closely left eye sample image of collection;
Based on the class Lis Hartel described closely human eye sample image of the self-adaptive enhancement algorithm training storehouse of seeking peace, obtain for the second cascade classifier of human eye detection closely;
Gather remote left eye sample image and remote right eye sample image, the described remote left eye sample image that will gather afterwards carries out flip horizontal, consist of remote human eye sample image storehouse by the remote left eye sample image after the upset and the described remote right eye sample image of collection, the described remote right eye sample image that perhaps will gather carries out flip horizontal, consists of remote human eye sample image storehouse by the remote right eye sample image after the upset and the described remote left eye sample image of collection;
Train described remote human eye sample image storehouse based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain the second cascade classifier for remote human eye detection.
4. human-eye positioning method as claimed in claim 3 is characterized in that, described closely left eye sample image includes closely people's left eye region on the face, and described closely right eye sample image includes closely people's right eye region on the face; Described remote left eye sample image includes remote people left eyebrow zone and left eye region on the face, and described remote right eye sample image includes remote people right eyebrow zone and right eye region on the face.
5. human-eye positioning method as claimed in claim 3, it is characterized in that, the first cascade classifier that described utilization is used for the detection of people's face extracts human face region from current image to be identified, according to the size of described human face region, select the step of corresponding described the second cascade classifier further to may further comprise the steps from the second cascade classifier two or more, that be used for the different distance human eye detection:
Utilize described the first cascade classifier from current image to be identified, to extract people's face rectangular area, and obtain height, width and the described people's face rectangular area positional information in described current image to be identified of described people's face rectangular area;
When the width of described people's face rectangular area during more than or equal to a predetermined width value, select described for the second cascade classifier of human eye detection closely, when the width of people's face rectangular area during less than described predetermined width value, select described the second cascade classifier for remote human eye detection.
6. human-eye positioning method as claimed in claim 3, it is characterized in that, described current image to be identified is video frame images, the first cascade classifier that described utilization is used for the detection of people's face extracts human face region from current image to be identified, according to the size of described human face region, select the step of corresponding described the second cascade classifier further to may further comprise the steps from the second cascade classifier two or more, that be used for the different distance human eye detection:
Utilize described the first cascade classifier from current image to be identified, to extract people's face rectangular area, and obtain height, width and the described people's face rectangular area positional information in described current image to be identified of described people's face rectangular area;
When the width of described people's face rectangular area during greater than a predetermined width value and a transition value sum, select described for the second cascade classifier of human eye detection closely, when the width of described people's face rectangular area during less than the difference of described predetermined width value and described transition value, select described the second cascade classifier for remote human eye detection, width when described people's face rectangular area is less than or equal to described predetermined width value and described transition value sum, and during more than or equal to the difference of described predetermined width value and described transition value, select selected the second cascade classifier of last video frame images.
7. human-eye positioning method as claimed in claim 1 is characterized in that, the step that described the second cascade classifier that described utilization is selected is located the position of human eye in the described current image to be identified further may further comprise the steps:
Utilize described the second cascade classifier of selecting, in the first half of the described human face region that extracts, extract the right eye rectangular area, obtain size and the positional information in described human face region thereof of described right eye rectangular area, and the coordinate of center on described facial image to be identified that obtains described right eye rectangular area according to size and the positional information in described human face region thereof of described right eye rectangular area, the coordinate of the center of described right eye rectangular area on described facial image to be identified is the right eye position in the described current image to be identified that the location obtains;
The first half of the described human face region that extracts is carried out flip horizontal, utilize described the second cascade classifier of selecting, extract the left eye rectangular area in the described human face region after upset, obtain the size of described left eye rectangular area and the positional information in the described human face region after upset thereof, and according to the size of described left eye rectangular area and the positional information in the described human face region after upset thereof, obtain the coordinate of center on unturned described facial image to be identified of described left eye rectangular area, the coordinate of the center of described left eye rectangular area on unturned described facial image to be identified is the left eye position of locating in the described current image to be identified that obtains.
8. a human eye positioning system is characterized in that, described system comprises:
The second cascade classifier selected cell, be used for utilizing the first cascade classifier that is used for the detection of people's face to extract human face region from current image to be identified, according to the size of described human face region, select corresponding the second cascade classifier from the second cascade classifier two or more, that be used for the different distance human eye detection;
Positioning unit, described the second cascade classifier that is used for utilizing described the second cascade classifier selected cell to select is located the position of human eye of described current image to be identified.
9. human eye positioning system as claimed in claim 7 is characterized in that, described system also comprises:
Training unit, be used for capturing sample image, and train described sample image based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, obtain the first cascade classifier of detecting for people's face, and two or more, be used for the second cascade classifier of different distance human eye detection.
10. human eye positioning system as claimed in claim 9 is characterized in that, described training unit comprises:
The first training module be used for to gather people's face sample image, trains described people's face sample image based on the class Lis Hartel self-adaptive enhancement algorithm of seeking peace, and obtains the first cascade classifier that detects for people's face;
Human eye sample image storehouse acquisition module closely, be used for gathering closely left eye sample image and closely right eye sample image, the described closely left eye sample image that will gather afterwards carries out flip horizontal, consist of closely human eye sample image storehouse by the closely left eye sample image after the upset and the described closely right eye sample image of collection, the described closely right eye sample image that perhaps will gather carries out flip horizontal, consists of closely human eye sample image storehouse by the closely right eye sample image after the upset and the described closely left eye sample image of collection;
The second training module is used for based on the class Lis Hartel described closely human eye sample image storehouse that the described closely human eye sample image of self-adaptive enhancement algorithm training storehouse acquisition module obtains of seeking peace, and obtains for the second cascade classifier of human eye detection closely;
Remote human eye sample image storehouse acquisition module, be used for gathering remote left eye sample image and remote right eye sample image, the described remote left eye sample image that will gather afterwards carries out flip horizontal, consist of remote human eye sample image storehouse by the remote left eye sample image after the upset and the described remote right eye sample image of collection, the described remote right eye sample image that perhaps will gather carries out flip horizontal, consists of remote human eye sample image storehouse by the remote right eye sample image after the upset and the described remote left eye sample image of collection;
The 3rd training module, the described remote human eye sample image storehouse for the self-adaptive enhancement algorithm of seeking peace based on the class Lis Hartel trains described remote human eye sample image storehouse acquisition module to obtain obtains the second cascade classifier for remote human eye detection.
11. human eye positioning system as claimed in claim 10 is characterized in that, described the second cascade classifier selects module to comprise:
Extraction module, described the first cascade classifier that is used for utilizing described training unit to obtain extracts people's face rectangular area from current image to be identified, and obtains height, width and the described people's face rectangular area positional information in described current image to be identified of described people's face rectangular area;
Select module, when being used for width when described people's face rectangular area that described extraction module obtains more than or equal to a predetermined width value, that selects that described training unit obtains is described for the second cascade classifier of human eye detection closely, when the width of described people's face rectangular area during less than described predetermined width value, described the second cascade classifier for remote human eye detection of selecting described training unit to obtain.
12. human eye positioning system as claimed in claim 10 is characterized in that, described the second cascade classifier selects module to comprise:
Extraction module, described the first cascade classifier that is used for utilizing described training unit to obtain extracts people's face rectangular area from current image to be identified, and obtains height, width and the described people's face rectangular area positional information in described current image to be identified of described people's face rectangular area;
Select module, when being used for width when described people's face rectangular area that described extraction module obtains greater than a predetermined width value and a transition value sum, that selects that described training unit obtains is described for the second cascade classifier of human eye detection closely, when the width of described people's face rectangular area that described extraction module obtains during less than the difference of described predetermined width value and described transition value, described the second cascade classifier for remote human eye detection of selecting described training unit to obtain, width when described people's face rectangular area that described extraction module obtains is less than or equal to described predetermined width value and described transition value sum, and during more than or equal to the difference of described predetermined width value and described transition value, select selected the second cascade classifier of last video frame images.
13. human eye positioning system as claimed in claim 8 is characterized in that, described positioning unit comprises:
Right eye position locating module, be used for utilizing described second cascade classifier of described the second cascade classifier selected cell selection, in the first half of the described human face region that described the second cascade classifier selected cell extracts, extract the right eye rectangular area, obtain size and the positional information in described human face region thereof of described right eye rectangular area, and the coordinate of center on described facial image to be identified that obtains described right eye rectangular area according to size and the described positional information in described human face region thereof of described right eye rectangular area, the coordinate of the center of described right eye rectangular area on described facial image to be identified is the right eye position in the described current image to be identified that described positioning unit location obtains;
Left eye position locating module, the first half for the described human face region that described the second cascade classifier selected cell is extracted is carried out flip horizontal, described the second cascade classifier that utilizes described the second cascade classifier selected cell to select, extract the left eye rectangular area in the described human face region after upset, obtain the size of described left eye rectangular area and the positional information in the described human face region after upset thereof, and according to the size of described left eye rectangular area and the described positional information in the described human face region after upset thereof, obtain the coordinate of center on unturned described facial image to be identified of described left eye rectangular area, the coordinate of the center of described left eye rectangular area on unturned described facial image to be identified is described positioning unit, left eye position in the described current image to be identified that the location obtains.
14. electronic equipment, comprise a display unit, it is characterized in that, described electronic equipment comprises also that just like each described human eye positioning system of claim 8 to 13 described display unit is for the described position of human eye that shows the described current image to be identified that described human eye positioning system location obtains.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210308876.3A CN102867172B (en) | 2012-08-27 | 2012-08-27 | A kind of human-eye positioning method, system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210308876.3A CN102867172B (en) | 2012-08-27 | 2012-08-27 | A kind of human-eye positioning method, system and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102867172A true CN102867172A (en) | 2013-01-09 |
| CN102867172B CN102867172B (en) | 2016-12-21 |
Family
ID=47446036
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210308876.3A Expired - Fee Related CN102867172B (en) | 2012-08-27 | 2012-08-27 | A kind of human-eye positioning method, system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102867172B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469018A (en) * | 2014-08-07 | 2016-04-06 | 中兴通讯股份有限公司 | Human eye positioning method and human eye positioning equipment |
| CN106203375A (en) * | 2016-07-20 | 2016-12-07 | 济南大学 | A kind of based on face in facial image with the pupil positioning method of human eye detection |
| WO2017114285A1 (en) * | 2015-12-30 | 2017-07-06 | 中国银联股份有限公司 | Eye recognition method and system |
| CN107072548A (en) * | 2014-11-03 | 2017-08-18 | 皇家飞利浦有限公司 | For people orientation and/or position automatic detection equipment, system and method |
| CN108460319A (en) * | 2017-02-22 | 2018-08-28 | 浙江宇视科技有限公司 | Abnormal face detecting method and device |
| WO2019205729A1 (en) * | 2018-04-26 | 2019-10-31 | 京东方科技集团股份有限公司 | Method used for identifying object, device and computer readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101350063A (en) * | 2008-09-03 | 2009-01-21 | 北京中星微电子有限公司 | Method and apparatus for locating human face characteristic point |
| CN101984453A (en) * | 2010-11-02 | 2011-03-09 | 中国科学技术大学 | Human eye recognition system and method |
| WO2011049550A1 (en) * | 2009-10-19 | 2011-04-28 | Hewlett-Packard Development Company, L.P. | Red eye detection aided by face detection |
| CN102096898A (en) * | 2009-11-17 | 2011-06-15 | 三星电子株式会社 | Method and apparatus for image processing |
| CN102339377A (en) * | 2010-07-21 | 2012-02-01 | 比亚迪股份有限公司 | Quick human-eye positioning method and device |
| CN102436578A (en) * | 2012-01-16 | 2012-05-02 | 宁波江丰生物信息技术有限公司 | Formation method for dog face characteristic detector as well as dog face detection method and device |
-
2012
- 2012-08-27 CN CN201210308876.3A patent/CN102867172B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101350063A (en) * | 2008-09-03 | 2009-01-21 | 北京中星微电子有限公司 | Method and apparatus for locating human face characteristic point |
| WO2011049550A1 (en) * | 2009-10-19 | 2011-04-28 | Hewlett-Packard Development Company, L.P. | Red eye detection aided by face detection |
| CN102096898A (en) * | 2009-11-17 | 2011-06-15 | 三星电子株式会社 | Method and apparatus for image processing |
| CN102339377A (en) * | 2010-07-21 | 2012-02-01 | 比亚迪股份有限公司 | Quick human-eye positioning method and device |
| CN101984453A (en) * | 2010-11-02 | 2011-03-09 | 中国科学技术大学 | Human eye recognition system and method |
| CN102436578A (en) * | 2012-01-16 | 2012-05-02 | 宁波江丰生物信息技术有限公司 | Formation method for dog face characteristic detector as well as dog face detection method and device |
Non-Patent Citations (2)
| Title |
|---|
| 王基帆，钱艺，童卫青: "基于adaboost算法的人眼定位", 《现代计算机（专业版）》 * |
| 马小强，符志鹏: "基于adaboost方法的实时人眼定位", 《电脑知识与技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469018A (en) * | 2014-08-07 | 2016-04-06 | 中兴通讯股份有限公司 | Human eye positioning method and human eye positioning equipment |
| CN105469018B (en) * | 2014-08-07 | 2020-03-13 | 中兴通讯股份有限公司 | Method and equipment for positioning human eyes |
| CN107072548A (en) * | 2014-11-03 | 2017-08-18 | 皇家飞利浦有限公司 | For people orientation and/or position automatic detection equipment, system and method |
| WO2017114285A1 (en) * | 2015-12-30 | 2017-07-06 | 中国银联股份有限公司 | Eye recognition method and system |
| CN106203375A (en) * | 2016-07-20 | 2016-12-07 | 济南大学 | A kind of based on face in facial image with the pupil positioning method of human eye detection |
| CN108460319A (en) * | 2017-02-22 | 2018-08-28 | 浙江宇视科技有限公司 | Abnormal face detecting method and device |
| WO2019205729A1 (en) * | 2018-04-26 | 2019-10-31 | 京东方科技集团股份有限公司 | Method used for identifying object, device and computer readable storage medium |
| US11093800B2 (en) | 2018-04-26 | 2021-08-17 | Boe Technology Group Co., Ltd. | Method and device for identifying object and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102867172B (en) | 2016-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110427905B (en) | Pedestrian tracking method, device and terminal | |
| US10127439B2 (en) | Object recognition method and apparatus | |
| CN102004899B (en) | Human face identifying system and method | |
| Yang et al. | Eigenjoints-based action recognition using naive-bayes-nearest-neighbor | |
| CN105940430B (en) | Personnel's method of counting and its device | |
| CN102867172A (en) | Human eye positioning method, system and electronic equipment | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| US10043064B2 (en) | Method and apparatus of detecting object using event-based sensor | |
| CN102547123B (en) | Self-adapting sightline tracking system and method based on face recognition technology | |
| Piątkowska et al. | Spatiotemporal multiple persons tracking using dynamic vision sensor | |
| CN101976330B (en) | Gesture recognition method and system | |
| KR20160088224A (en) | Method for recognizing an object and apparatus thereof | |
| CN102385703B (en) | A kind of identity identifying method based on face and system | |
| Demiröz et al. | Feature-based tracking on a multi-omnidirectional camera dataset | |
| KR102595604B1 (en) | Method and apparatus of detecting object using event-based sensor | |
| Taha et al. | Human activity recognition for surveillance applications | |
| CN101551852B (en) | Training system, training method and detection method | |
| JP5754990B2 (en) | Information processing apparatus, information processing method, and program | |
| CN106919891B (en) | A kind of image processing method and device | |
| CN101655910A (en) | Training system, training method and detection method | |
| CN105812942A (en) | Data interaction method and device | |
| CN109508576A (en) | A kind of abnormal driving behavioral value method, apparatus and electronic equipment | |
| Liu et al. | Spatial-temporal motion information integration for action detection and recognition in non-static background | |
| Stalder et al. | Exploring context to learn scene specific object detectors | |
| CN107111363B (en) | Method, device and system for monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20161221 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |