Apparatus and method for moving object detection

• US 8,447,069 B2
• Filed: 02/09/2010
• Issued: 05/21/2013
• Est. Priority Date: 12/10/2009
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Huang, Chung-Hsien, Chou, Cheng-Chuan, Wu, Yi-Ta, Kao, Chao-Hung, Shih, Ming-Yu
• Primary Examiner(s): Wang, Claire X
• Application Number:
  US12/702,312
• Publication Number:
  US 20110142283A1
• Time in Patent Office: 1,197 Days
• Field of Search: None
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/20221 Image fusion; Image merging
• CPC Class Codes: G06T 2207/20224 Image subtraction
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06T 7/254 involving subtraction of images

Abstract

An apparatus and method for moving object detection computes a corresponding frame difference for every two successive image frames of a moving object, and segments a current image frame of the two successive image frames into a plurality of homogeneous regions. At least a candidate region is further detected from the plurality of homogeneous regions. The system gradually merges the computed frame differences via a morphing-based technology and fuses the merged frame difference with the at least a candidate region, thereby obtaining the location and a complete outline of the moving object.

Claims

1. An apparatus for moving object detection, comprising:

a frame difference module, for each two successive images of a moving object, computing a frame difference corresponding to said two successive images, said two successive images being a current image and a previous image respectively;

a candidate search module, segmenting inputted said current image into a plurality of homogeneous regions, determining a plurality of uniform blocks from said plurality of homogeneous regions, and then detecting at least a candidate region from said plurality of uniform blocks; and

a moving object fusion module, fusing a plurality of frame differences with said at least a candidate region, aided by a gradual morphing-based technique to obtain a location and a complete outline of said moving object.

2. The apparatus as claimed in claim 1, wherein said moving object fusion module further includes:

a morphing module, said morphing module selecting a candidate region difference image mask of said at least a candidate region of said moving object, and merging said plurality of frame differences into a single object through said gradual morphing-based technique; and

a fusing module, said fusing module fusing said at least a candidate region and said single object generated by said morphing module, thereby obtaining said complete outline of said moving object.

3. The apparatus as claimed in claim 1, wherein two types of regions are selected from said plurality of uniform blocks;

one type is regions included inside said plurality of uniform blocks, and the other type is regions in surrounding areas of said plurality of uniform blocks connecting to image edges.

4. The apparatus as claimed in claim 1, wherein said moving object fusion module generates a plurality of successive and smooth intermediate images between a source image and a destination image through said gradual morphing-based technique.

5. The apparatus as claimed in claim 4, wherein said complete outline of said moving object is a union result of said source image and all intermediate images of said gradual morphing-based technique.

6. The apparatus as claimed in claim 1, wherein said frame difference generated by said frame difference module is an accumulated difference image accumulated by a plurality of difference images.

7. The apparatus as claimed in claim 1, wherein said apparatus further includes a moving object tracking module to predict, compensate and update said location of said moving object.

8. The apparatus as claimed in claim 7, wherein said moving object tracking module continuously predicts and updates a state of said moving object to achieve moving object tracking, through a motion model and an observation model of said moving object.

9. The apparatus as claimed in claim 7, wherein said moving object tracking module maintains a tracker, and said tracker has a moving object list at a current time and a moving object hypothesis list having been tracked at a previous time.

10. The apparatus as claimed in claim 1, wherein said apparatus further includes an image alignment module, and when said two successive images are captured from said moving object in motion, said image alignment module first aligns said previous image to said current image, and then said frame difference module computes said frame difference of the aligned said previous image and said current image.

11. A method for moving object detection, applicable to an apparatus for moving object detection, said method comprising:

for each two successive images of a moving object, computing a frame difference corresponding to said two successive images, said two successive images being a current image and a previous image respectively;

segmenting said current image into a plurality of homogeneous regions, determining a plurality of uniform blocks from said plurality of homogeneous regions, and then detecting at least a candidate region from said plurality of uniform blocks; and

fusing a plurality of frame differences with said at least a candidate region through a moving object fusion module, and aided by a gradual morphing-based technology to obtain a location and a complete outline of said moving object.

12. The method as claimed in claim 11, wherein said fusing a plurality of frame differences with said at least a candidate region, and aided by a gradual morphing-based technology further includes:

based on information of at least an accumulated difference image and at least a segmentation label, finding a difference image set falling inside a candidate region where said at least a segmentation label belongs to;

if a number of difference images in said difference image set is greater than 1, selecting two most neighboring difference images from said difference image set and using a morphing technique to merge said two most neighboring difference images into another difference image, adding said another difference image to said difference image set, and obtaining intersection of said difference image set and said candidate region; and

repeating above steps until all inputted segmentation labels corresponding to all of said plurality of homogeneous regions are processed.

13. The method as claimed in claim 11, wherein said detecting at least a candidate region from said plurality of uniform blocks further includes:

checking each of said plurality of uniform blocks, observing whether different regions exist inside said plurality of uniform blocks, or regions in surrounding of said plurality of uniform blocks connecting image edges to find said at least a candidate region.

14. The method as claimed in claim 11, wherein said frame difference is an accumulated difference image accumulated by a plurality of difference images in a weighted manner.

15. The method as claimed in claim 12, wherein said morphing technique generates a plurality of successive and smooth intermediate images between a source image and a destination image by using an interpolation method.

16. The method as claimed in claim 13, wherein a ratio of a contour distance and a real distance of a region in said surrounding of said plurality of uniform blocks connecting an image edge is higher than a threshold.

17. The method as claimed in claim 11, wherein said method further including:

through a motion model and an observation model, continuously predicting and updating a state of said moving object for further tracking said moving object.

18. The method as claimed in claim 17, wherein said moving object tracking further includes:

inputting a moving object hypothesis list tracked at a plurality of previous times, and predicting a state at a current time via a filter;

for an inputted image at said current time, obtaining a moving object list at said current time by said location and said complete outline of said moving object;

constructing a correlation matrix to describe a spatial correlation between elements of said moving object hypothesis and said moving object lists;

performing summation on rows and columns of said correlation matrix, and performing corresponding event processing according to an output value of said summation; and

based on said corresponding event processing, updating said moving object hypothesis list.

19. The method as claimed in claim 18, wherein said moving object hypothesis list records a plurality of moving object hypotheses tracked at said plurality of previous times, and said moving object list records a plurality of foreground object rectangles at said current time.

20. The method as claimed in claim 18, wherein said output value of said summation of rows and columns of said correlation matrix reflects one of events at least including addition, disappearing, steady tracking of a corresponding moving object, and merging and splitting among a plurality of moving objects.

21. A computer program product, comprising:

at least a program storage formed by non-transitory computer-readable medium, readable by a digital processing apparatus; and

a program of a plurality of instructions, tangibly embodied in said at least a program storage and executable by said digital processing apparatus to perform a moving object detection method, said method comprising;

for each two successive images of a moving object, computing a frame difference corresponding to said two successive images, said two successive images being a current image and a previous image respectively;

segmenting said current image into a plurality of homogeneous regions, determining a plurality of uniform blocks from said plurality of homogeneous regions, and then detecting at least a candidate region from said plurality of uniform blocks; and

fusing a plurality of said frame differences with said at least a candidate region through a moving object fusion module, and aided by a gradual morphing-based technology to obtain a location and a complete outline of said moving object.

22. The computer program product as claimed in claim 21, wherein said computer program product is further connected to a moving object tracking module for predicting, compensating and updating said location of said moving object.

23. The computer program product as claimed in claim 21, wherein said moving object detection method further includes:

through a motion model and an observation model, continuously predicting and updating a state of said moving object for further tracking said moving object.