Summary of the invention
The embodiment of the present invention provides a kind of method for supervising and device of cluttered scenes, can monitoring image to obtain the scene that whether causes confusion in the image that equipment obtains, and in the time causing confusion scene, notifies in time user.
The technical scheme that the embodiment of the present invention provides is as follows:
A method for supervising for cluttered scenes, comprising: in the image of image capture apparatus picked-up, detect target congested area; The target congested area detecting is divided into image block; And the motion vector of definite each image block marking off; According to the motion vector of each image block of determining, determine the scene that whether causes confusion in target congested area; And determining while there is cluttered scenes, send the notice causing confusion in the image of image capture apparatus picked-up.
A supervising device for cluttered scenes, comprising:
Detecting unit, for detecting target congested area at the image of image capture apparatus picked-up;
Division unit, is divided into image block for the target congested area that detecting unit is detected;
The first determining unit, the motion vector of each image block marking off for definite division unit;
The second determining unit, for the motion vector of each image block of determining according to the first determining unit, determines the scene that whether causes confusion in target congested area;
Notification unit, while there is cluttered scenes, sends the notice causing confusion in the image of image capture apparatus picked-up for determining in the second determining unit.
The embodiment of the present invention by detecting target congested area in the image of image capture apparatus picked-up, the target congested area detecting is divided into image block, and the motion vector of definite each image block marking off, according to the motion vector of each image block of determining, determine the scene that whether causes confusion in target congested area; Determining while there is cluttered scenes, notify monitor staff, the scene that whether causes confusion in can monitoring image is provided, and in the time causing confusion scene, has notified in time monitor staff's possible technique scheme.
Embodiment
Prior art adopts the method for supervising of manual observation monitor, has the problem of easily missing or ignoring the serious congested conditions occurring in part watch-dog wherein.The embodiment of the present invention is by detecting personnel's congested area the image first obtaining from image acquisition equipment, and further determine whether to have occurred cluttered scenes according to the motion vector of each several part in congested area, solve the problem that detects cluttered scenes the image that prior art cannot obtain from image acquisition equipment and notify user, feasible technical scheme is provided.
Below in conjunction with each accompanying drawing, embodiment of the present invention technical scheme main realized to principle, embodiment and the beneficial effect that should be able to reach is explained in detail.The embodiment of the present invention is introduced the method for supervising of cluttered scenes that the present invention proposes as an example of human body head image example, the method that the embodiment of the present invention proposes is equally applicable to monitor other targets (such as automobile, animal etc.) scene that whether causes confusion.
Please refer to accompanying drawing 1, it is as follows that the embodiment of the present invention main realized principle process:
Step 101, the grader based in Data Mining and clustering method, and the first predetermined threshold, determine personnel congested area the image to be detected getting from image acquisition equipment;
Step 102, personnel's congested area that step 101 is obtained is divided into the image block of preliminary dimension, and based on piece base method for estimating, determines the motion vector of each image block;
Step 103, the motion vector of each image block of determining based on step 102, determines in congested area, whether to have occurred cluttered scenes;
Step 104, determines while causing confusion scene in described congested area in step 103, notifies user.
Below will be according to foregoing invention principle of the present invention, introduce in detail an embodiment the main principle that realizes of the inventive method is explained in detail and is illustrated.
Please refer to accompanying drawing 2, the image to be detected absorbing for the first-class image capture apparatus of making a video recording.Sorting technique based in Data Mining and clustering method, and the first predetermined threshold, determine congested area the image absorbing from image capture apparatus, and detailed process please refer to accompanying drawing 3:
Step 301 first detects the position at human body predetermined position place from image to be detected, and the present embodiment is introduced as an example of head example.Treat detected image with predetermined ratio and carry out continuous scaling, the image series then obtaining after to scaling with the window of preliminary dimension and displacement step-length travels through.In ergodic process, utilize the two class graders that go out according to the sample training of a large amount of head image samples and non-head image in advance to differentiate whether the determined image-region of preliminary dimension window after each displacement is human body head image.Accordingly, please refer to accompanying drawing 4, can obtain the information such as position, size of all human body heads that detect in image to be detected.The accuracy detecting in order to improve head image, obtaining after the information such as position, size of human body head in the image to be detected shown in accompanying drawing 4, can also such as, be further analyzed the image of the human body head detecting in conjunction with human body head inherent feature (feature such as ear, hair), be respectively similarity corresponding to each human body head image calculation, filter out human body head image that similarity is high as the human body head image in the image to be detected detecting.
Step 302, the clustering methods such as the k-means method of utilization based in Data Mining, the positional information at the human body head place detecting from image to be detected according to above-mentioned steps 301 is obtained alternative congested area from image to be detected.The groundwork principle of k-means method is as follows: (k < is n) individual as initial cluster center in the n from image to be detected human body head image, to select k, each human body head image in definite all n-k residue human body head images and the distance between above-mentioned each initial cluster center respectively, each residue human body head image is distributed to cluster centre the shortest with this human body head distance in initial cluster center, then calculate the average (each human body head image and the mean value as the distance between the human body head image of initial cluster center in this cluster) that adds each cluster after new human body head image, repeat said process until canonical measure function (conventionally using the variance of each human body head in this cluster and the distance between the human body head of initial cluster center as the canonical measure function) convergence of each cluster.Using the rectangular image region, border at each cluster place as alternative congested area.
Step 303, the number of the human body head image comprising in the alternative congested area of obtaining in step 302 (is for example greater than to the first predetermined threshold, the first predetermined threshold is 3 in the present embodiment) alternative congested area as the congested area of determining, please refer to accompanying drawing 5, is the position view of congested area in image to be detected.
Please refer to accompanying drawing 6A, for above-mentioned personnel's congested area of obtaining being carried out to the schematic diagram of piecemeal.For example, according to the window of preliminary dimension (8 pixel * 8 pixels), personnel's congested area of above-mentioned acquisition is carried out to piecemeal processing, and determine the motion vector of each image block.Introduce as an example of the motion vector of determining the image block A in the image to be detected in accompanying drawing 5A example the detailed process of determining each image block motion vector below:
Please refer to accompanying drawing 6B, with above-mentioned preliminary dimension, for example, in the reference frame former frame image of image to be detected (with respect to), take this image block, the position in image to be detected is as datum mark, for example, image in predetermined investigative range (40 pixel * 40 pixels) carries out piecemeal; Determine image block the highest with this image block A similarity in investigative range, as shown in B in accompanying drawing 6B; The motion vector of further determining image block B and this image block A, as shown in the arrow in accompanying drawing 6B, motion vector length attribute has shown the distance of the object motion that image block is corresponding, motion vector direction attribute list understands the direction of the target travel that image block is corresponding.
Obtaining after the motion vector of each image block in congested area, choose the image block that motion vector length attribute value is greater than the second predetermined threshold, when the number that is greater than the image block of the second predetermined threshold when motion vector length attribute value is greater than the 3rd predetermined threshold, determine that according to the following equation motion vector length attribute value is greater than the variance sum of motion vector direction property value between each image block of the second predetermined threshold:
Wherein n represents that motion vector length attribute is greater than the number of the image block of the second predetermined threshold, X
irepresent that motion vector length attribute is greater than the motion vector direction property value of i image block of the second predetermined threshold, span is 0 °~360 °, and X represents that motion vector length attribute is greater than the motion vector direction attribute average of each image block of the second predetermined threshold.
In the time that variance sum σ is greater than the 4th predetermined threshold, the object motion speed that each image block is corresponding is described, and the direction of motion has obvious difference, judge and in congested area, occurred confusion phenomena.
Cause confusion after phenomenon judging in congested area, can adopt the mode such as the tinkle of bells, light to send alarm to user, reminding user should take corresponding measure to stop to reduce the loss.
The technical scheme that the embodiment of the present invention proposes, the first grader based in Data Mining and clustering method, and predetermined threshold, determine congested area the image to be detected getting from image acquisition equipment; Then the window based on preliminary dimension carries out piecemeal to congested area, and based on piece base method for estimating, determines the motion vector of each image block obtaining after piecemeal; According to the length attribute value of the motion vector of each image block of determining and direction property value, at object motion speed corresponding to each image block, and when the direction of motion has obvious difference, judge and in congested area, occurred confusion phenomena, thereby determine in image to be detected and occurred cluttered scenes.Adopt technique scheme, can detect in image to be detected, whether to have occurred cluttered scenes, thereby solved prior art and cannot from image, detect the problem of cluttered scenes.
Correspondingly, the embodiment of the present invention also provides a kind of supervising device of cluttered scenes, please refer to accompanying drawing 7, and this device comprises detecting unit 701, division unit 702, the first determining unit 703, the second determining unit 704 and notification unit 705, wherein,
Detecting unit 701, for detecting target congested area at the image of image capture apparatus picked-up;
Division unit 702, for according to preliminary dimension, the target congested area that detecting unit 701 is detected is divided into image block;
The first determining unit 703, the motion vector of each image block marking off for definite division unit 702, specifically can adopt based on piece base estimation mode, determines the motion vector of each image block
The second determining unit 704, for the motion vector of each image block of determining according to the first determining unit 703, determines the scene that whether causes confusion in target congested area;
Notification unit 705, while there is cluttered scenes, sends the notice causing confusion in the image of image capture apparatus picked-up for determining in the second determining unit 704.
Preferably, please refer to accompanying drawing 8, above-mentioned detecting unit specifically comprises detection sub-unit 801, first definite subelement 802, second definite subelement 803, wherein,
Detection sub-unit 801, for detecting the image of target privileged site at the image of image capture apparatus picked-up;
First determines subelement 802, for the position of the target privileged site image that detects according to detection sub-unit 801, determines at least one the alternative congested area existing in the image of image capture apparatus picked-up;
Second determines subelement 803, for determining the alternative congested area determined of subelement 802 first, will comprise alternative congested area that target privileged site picture number is greater than the first predetermined threshold as the congested area detecting.
Preferably, please refer to accompanying drawing 9, the second determining unit in accompanying drawing 7 specifically comprises the 3rd definite subelement 901, obtains subelement 902 and the 4th definite subelement 903, wherein,
The 3rd determines subelement 901, for according to the motion vector of each image block of determining, determines that motion vector length attribute value is greater than the quantity of the image block of the second predetermined threshold;
Obtain subelement 902, for in the time that the 3rd quantity of determining the image block determined of subelement 901 is greater than the 3rd predetermined threshold, obtain the variance sum of the motion vector direction property value between each image block that motion vector length attribute value is greater than the second predetermined threshold;
The 4th determines subelement 903, in the time obtaining variance sum that subelement 902 obtains and be greater than the 4th predetermined threshold, determines in target congested area and has occurred cluttered scenes.
Obviously, those skilled in the art can carry out various changes and modification and not depart from the spirit and scope of the present invention the present invention.Like this, if within of the present invention these are revised and modification belongs to the scope of the claims in the present invention and equivalent technologies thereof, the present invention is also intended to comprise these changes and modification interior.