Method for moving targets tracking and number counting

• US 8,213,679 B2
• Filed: 07/24/2009
• Issued: 07/03/2012
• Est. Priority Date: 01/25/2007
• Status: Expired due to Fees

• Current Assignee: Shanghai YAO WEI Industry Company Limited
• Original Assignee: Shanghai YAO WEI Industry Company Limited
• Inventors: Yao, Wei
• Primary Examiner(s): Johns, Andrew W
• Assistant Examiner(s): GORADIA, SHEFALI DINESH
• Application Number:
  US12/509,106
• Publication Number:
  US 20100021009A1
• Time in Patent Office: 1,075 Days
• Field of Search: 382/100, 382/103, 382/167, 382/173, 382/177, 348/169, 348/208.14, 348/208.6
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30241 Trajectory
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/20 Analysis of motion motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats

Abstract

The invention discloses a method for moving targets tracking and number counting, comprising the steps of: a). acquiring continuously the video images comprising moving targets; b). acquiring the video image of a current frame, and pre-processing the video image of the current frame; c). segmenting the target region of the processed image, and extracting the target region; d). matching the target region of the current frame obtained in step c) with that of the previous frame based on an online feature selection to establish a match tracking link; and e). determining the number of the targets corresponding to each match tracking link based on the target region tracks recorded by the match tracking link. The invention can solve the problem of low precision of the number statistic results caused by the bad environment, such as that the distribution of the illumination is extremely not equilibrium spatially, the change in a time period is complicated, the change of the gesture during the people goes by is evident, and the like, under the normal application condition.

Claims

1. A method for moving targets tracking and number counting, comprising the steps of,

a). acquiring continuously video images comprising moving targets;

b). acquiring the video image of a current frame, and pre-processing the video image of the current frame;

c). performing target region segmenting upon the pre-processed image, and extracting a target region;

d). matching the target region of the current frame obtained in step c) with the target region of the previous frame based on an online feature selection to establish a match tracking link; and

e). determining the number of the targets corresponding to each match tracking link based on the target region tracks recorded by the match tracking link;

f). acquiring the video images from step a) for each frame, performing step b)-step c), and accumulating the number of the targets corresponding to each tracking link,

wherein the step of segmenting the target region in step c) comprising;

c1). segmenting the pre-processed image into a plurality of separated first regions;

c2). computing the motion information of each of the first regions;

c3). determining and extracting the motion region based on the motion information;

c4). verifying whether each of the motion regions is a target region or not; and

c5). extracting the verified target region.

2. The method for moving targets tracking and number counting of claim 1, wherein in said step c1), the step of segmenting said image comprising:

c11). segmenting the image into a plurality of second regions;

c12). computing the motion information of each of the second regions;

c13). computing the descriptive attribute information of each of the second regions;

c14). creating a region neighboring diagram based on the spatial position relationships among the second regions and the descriptive attribute information; and

c15). combining the second regions into a first region based on the descriptive attribute information and the region neighboring diagram.

3. The method for moving targets tracking and number counting of claim 2, wherein in said step ell), the method for segmenting the image into a plurality of second regions comprises watershed segmentation algorithm and seed point region growing algorithm.

4. The method for moving targets tracking and number counting of claim 2, wherein in said step c11), the internal pixels in each of the second regions possess uniform pixel attribute, the pixel attribute comprising color, gradient strength, whether an edge point or not, and edge response strength.

5. The method for moving targets tracking and number counting of claim 2, wherein in said step c12, the motion information in the second region comprises motion vector of the second region and the motion probability value of the edge pixels.

6. The method for moving targets tracking and number counting of claim 2, wherein in said step c13, the descriptive attribute information comprises:

first type attributes for characterizing the internal attribute of each second region comprising color distribution attribute, grey scale distribution attribute, and motion information attribute;

second type attributes for characterizing the difference between any two spatially adjacent second regions comprising a difference attribute of the grey scale mean values, a difference attribute of the color mean values, and a difference attribute of the motion vectors, between the second regions; and

third type attributes for characterizing a common edge feature between any two spatially adjacent second regions comprising gradient strength attribute, the number of the edge points, and the motion probability attribute of the common edge points.

7. The method for moving targets tracking and number counting of claim 6, wherein in said step c14), the step of creating a region neighboring diagram further comprises:

representing each of the second regions with a node, and storing the information of the first type attributes in the node;

creating a bi-directional connecting arc between the nodes corresponding to two spatially adjacent second regions, and storing the information of the second type and the third type attributes in the connecting arc; and

obtaining a region neighboring diagram of the bi-directional connecting arc comprising a plurality of the nodes and the connecting nodes.

8. The method for moving targets tracking and number counting of claim 6, wherein in said step c15), the step of combining the second regions into the first region comprises:

under a restriction condition that the similarity of the second type attributes, the third type attributes reaches to a predetermined value, combining the second regions in a sequence according to the similarity of the first type attributes from large to small, until the similarity of first type attributes is smaller than a predetermined value;

under a restriction condition that the similarity of the first type attributes, the third type attributes reaches to a predetermined value, combining the second regions in a sequence according to the similarity second type attributes from large to small, until the similarity of the second type attributes is smaller than a predetermined value; and

under a restriction condition that the similarity of the first type attributes, the second type attributes reaches to a predetermined value, combining the second regions in a sequence according to the similarity of the third type attributes from large to small, until the similarity of the third type attributes is smaller than a predetermined value.

9. The method for moving targets tracking and number counting of claim 1, wherein in said step c4), the step of verifying whether each of the motion regions is target region or not uses a elastic matching method, wherein said elastic matching method comprises:

a training stage step comprising;

selecting a set of image regions of a quantity TrainSum which comprise target regions as a training image set;

describing the training image being represented by two dimensional array Accm[x] [y], wherein x and y in the array are the horizontal coordinate and vertical coordinate of the pixels in the training image set, respectively;

for each point P(x,y) in each training image, if it is located inside the target region, increasing the value of the element Accm[x][y] corresponding to two dimensional array Aecm by 1; and

a motion region classifying step comprising;

setting an accumulator A for each of the motion regions, the initial value of which set as A=0;

for each of all of the pixels Q(x, y) in the motion region, adding the value of the numerical array element Accm[x][y] corresponding to the pixel to the accumulator A to obtain an accumulated value;

if the accumulated value>

TrainSum*S*Threshold, and the total number of the pixels S conforms to Smin<

S<

Smax, determining said region as a target region, otherwise, determining said region not as a target region;

wherein the parameter Threshold indicates a critical coefficient for allowing to confirm a motion region being a target region, and Smin and Smax indicate the lower limit and the upper limit of the total number of the pixels in the target region, respectively.

10. The method for moving targets tracking and number counting of claim 1, wherein said step d) further comprising:

d1. establishing a feature pool comprising a plurality of target features;

d2. for at least part of the features in the feature pool, computing the match degree between each of the target features of the respective target region in the current frame and each of the target features of the respective target region in the previous frame;

d3). computing a synthetic match degree measure between the respective target region in the current frame and the respective target region in the previous frame based on the match degrees;

d4). establishing a match mapping relationship between a target region in the current frame and a previous frame'"'"'s target region having the maximum synthetic match degree measure with respect to that in the current frame; and

d5. establishing a match tracking link comprising the target region in the current frame based on the match mapping relationship.

11. The method for moving targets tracking and number counting of claim 10, wherein in said step d3), the step of computing synthetic match degree measure comprises:

selecting a predetermined proportion of features in the feature pool, for which the match degree between the target region in the current frame and the target region in the previous frame are the highest, performing a weighted computation thereon, and obtaining the synthetic match degree measure between said two target regions.

12. The method for moving targets tracking and number counting of claim 10, wherein in said step d5), the step of establishing the match tracking link comprises:

finding the tracking link corresponding to the target region in the previous frame based on the match mapping relationship of step d4), and adding the target region in the current frame to the tracking link corresponding to the target region in the previous frame; and

if the corresponding tracking link cannot be found in the previous frame, then establishing a new tracking link.

13. The method for moving targets tracking and number counting claim 1, wherein said step e) further comprises:

for each not counted tracking link, deciding whether it is a tracking link suitable for counting based on the spatial displacement track of the target and the variation rule of the target size over time; and

for each counted tracking link, verifying the accuracy of the counting result of the tracking link, if the previous counting of the tracking link is not consistent with the current counting, then using the current counting, meanwhile, modifying the previous count result, until the tracking link leaves entirely from the image acquisition region.