Object recognition method and object recognition apparatus using the same

• US 9,400,924 B2
• Filed: 12/12/2014
• Issued: 07/26/2016
• Est. Priority Date: 05/23/2014
• Status: Expired due to Fees

• Current Assignee: We Jump Ar Co., Ltd.
• Original Assignee: Industrial Technology Research Institute
• Inventors: Yu, Chih-Hsiang, Yang Mao, Shys-Fan, Peng, Wen-Wei, Jeng, Tzuan-Ren, Lin, Hsien-Chang
• Primary Examiner(s): Bitar, Nancy
• Application Number:
  US14/569,198
• Publication Number:
  US 20150339520A1
• Time in Patent Office: 592 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
• CPC Class Codes: G06V 40/28 Recognition of hand or arm movements, e.g. recognition of deaf sign language static hand signs G06V40/113

Abstract

An object recognition method and an object recognition apparatus using the same are provided. In one or more embodiments, a real-time image including a first object is acquired, and a chamfer distance transform is performed on the first object of the real-time image to produce a chamfer image including a first modified object. Preset image templates each including a second object are acquired, and the chamfer distance transform is performed on the second object of each preset image template to produce a chamfer template including a second modified object. When the difference between the first modified object and the second modified object is less than a first preset error threshold, the object recognition apparatus may operate according to a control command corresponding to the preset image template.

Claims

1. An object recognition method, applied to an object recognition apparatus and comprising:

acquiring a real-time image comprising a first object;

performing a chamfer distance transform on the first object of the real-time image to produce a chamfer image comprising a first modified object;

acquiring a plurality of preset image templates comprising a second object;

performing the chamfer distance transform on the second object of each of the plurality of preset image templates to produce a chamfer template comprising a second modified object;

determining whether difference between the first modified object and one of the second modified objects is less than a first preset error threshold;

when the difference between the first modified object and the second modified object is less than the first preset error threshold, looking up a control command according to the preset image template corresponding to the second modified object, the object recognition apparatus operating according to the control command;

comparing a size or location of the first object with sizes or locations of the second objects to determine whether to adjust the preset image templates, comprising;

detecting a first centroid of the first object;

detecting a first maximum inscribed circle according to the first centroid of the first object;

detecting a second centroid of each of the second objects;

detecting second maximum inscribed circles according to the second centroids of the second objects respectively;

determining whether a size or location of the first maximum inscribed circle is equal to a size or location of one of the second maximum inscribed circles; and

adjusting the preset image template comprising the second maximum inscribed circle when the size or location of the first maximum inscribed circle is different from the size or location of the second maximum inscribed circle; and

when the size or location of the first object is different from the size or location of one of the second object, adjusting the preset image template comprising the second object.

2. The object recognition method according to claim 1, wherein the step of adjusting the preset image template comprises scaling, shifting, or rotating the second object of the preset image template.

3. The object recognition method according to claim 1, further comprising:

comparing a size or location of the first modified object with sizes or locations of the second modified objects to determine whether to adjust the chamfer templates; and

when the size or location of the first modified object is different from the size or location of one of the second modified objects, adjusting the chamfer template comprising the second modified object.

4. The object recognition method according to claim 3, wherein the step of comparing the size or location of the first modified object with the sizes or locations of the second modified objects to determine whether to adjust the chamfer templates comprises:

detecting a third centroid of the first modified object;

detecting a third maximum inscribed circle according to the third centroid of the first modified object;

detecting a fourth centroid of each of the second modified objects;

detecting fourth maximum inscribed circles according to the fourth centroids of the second modified objects respectively;

determining whether a size or location of the third maximum inscribed circle is equal to a size or location of each of the fourth maximum inscribed circles; and

adjusting the chamfer template comprising the fourth maximum inscribed circle when the size or location of the third maximum inscribed circle is different from the size or location of one of the fourth maximum inscribed circles.

5. The object recognition method according to claim 1, wherein the chamfer distance transform is performed on edges or skeletons of the first object and the second object.

6. The object recognition method according to claim 1, wherein when movement of the first object in a sequence of next real-time images matches a preset trace model, the control command is sent.

7. An object recognition apparatus, comprising:

a first image capturing device, for recording a real-time image comprising a first object;

a storage device, for storing preset image templates each of which comprises a second object; and

a processing device connected to the first image capturing device and the storage device, for receiving the real-time image from the first image capturing device, receiving the preset image templates from the storage device, and performing the following steps;

performing a chamfer distance transform on the first object of the real-time image to produce a chamfer image comprising a first modified object;

respectively performing the chamfer distance transform on the second objects of the preset image templates to produce chamfer templates each of which comprises a second modified object;

determining whether difference between the first modified object and one of the second modified objects is less than a first preset error threshold;

when the difference between the first modified object and the second modified object is less than the first preset error threshold, looking up a control command according to the preset image template corresponding to the second modified object, the object recognition apparatus operating according to the control command;

comparing a size or location of the first object with a size or location of each of the second objects to determine whether to adjust the preset image templates, and when the size or location of the first object is different from the size or location of one of the second objects, the processing device adjusting the preset image template comprising the second object; and

detecting a first centroid of the first object, detecting a first maximum inscribed circle in the first object according to the first centroid, detecting a second centroid of each of the second objects, detecting a second maximum inscribed circle in each of the second objects according to each of the second centroids, and determining whether a size or location of the first maximum inscribed circle is equal to a size or location of each of the second maximum inscribed circles; and

when the size or location of the first maximum inscribed circle is different from the size or location of one of the second maximum inscribed circles, the processing device adjusting the preset image template comprising the second maximum inscribed circle.

8. The object recognition apparatus according to claim 7, wherein when adjusting the preset image template, the processing device scales, shifts, or rotates the second object of the preset image template.

9. The object recognition apparatus according to claim 7, wherein the processing device further compares a size or location of the first modified object with sizes or locations of the second modified objects to determine whether to adjust the chamfer templates, and when the size or location of the first modified object is different from the size or location of one of the second modified objects, the processing device adjusts the chamfer template comprising the second modified object.

10. The object recognition apparatus according to claim 9, wherein the processing device detects a third centroid of the first modified object, detects a third maximum inscribed circle according to the third centroid of the first modified object, detects a fourth centroid of each of the second modified objects, detects fourth maximum inscribed circles according to the fourth centroids of the second modified objects respectively, and determines whether a size or location of the third maximum inscribed circle is equal to a size or location of each of the fourth maximum inscribed circles;

and when the size or location of the third maximum inscribed circle is different from the size or location of one of the fourth maximum inscribed circles, the processing device adjusts the chamfer template comprising the fourth maximum inscribed circle.

11. The object recognition apparatus according to claim 7, wherein the chamfer distance transform is performed on edges or skeletons of the first object and the second objects.

12. The object recognition apparatus according to claim 7, wherein when movement of the first object in a sequence of next real-time images matches a preset trace model, the processing device sends the control command.

13. The object recognition apparatus according to claim 12, wherein the first image capturing device has a field of view (FOV) comprising a preset sensing region, the preset sensing region is smaller than or equal to the FOV, and when the first object of the sequence of next real-time images moves out of the preset sensing region, the control command is sent to control the first image capturing device to perform image capturing.

14. The object recognition apparatus according to claim 12, wherein the first image capturing device has a first FOV, the object recognition apparatus further comprises a second image capturing device having a second FOV, the first FOV is larger than or equal to the second FOV, and when the first object of the sequence of next real-time images from the second FOV to the first FOV, the control command is sent to control the second image capturing device to perform image capturing.

15. An object recognition method, applied to an object recognition apparatus and comprising:

acquiring an original frame from a first image capturing device;

performing an image pre-processing procedure on the original frame to produce a real-time image comprising a first object;

performing a chamfer distance transform on the first object of the real-time image to generate a chamfer image comprising a first modified object;

acquiring a plurality of preset image templates comprising a second object;

performing the chamfer distance transform on the second object of each of the plurality of preset image templates to generate a chamfer template comprising a second modified object;

determining whether difference between the first modified object and one of the second modified objects is less than a first preset error threshold;

when the difference between the first modified object and one of the second modified objects is less than the first preset error threshold, looking up a control command according to the preset image template corresponding to the second modified object, the object recognition apparatus operating according to the control command;

comparing a size or location of the first object with sizes or locations of the second objects to determine whether to adjust the preset image templates, comprising;

detecting a first centroid of the first object;

detecting a first maximum inscribed circle in the first object according to the first centroid;

detecting a second centroid of each of the second objects;

detecting second maximum inscribed circles in the second objects according to the second centroids respectively;

determining whether a size or location of the first maximum inscribed circle is equal to a size or location of each of the second maximum inscribed circles; and

adjusting the preset image template comprising the second maximum inscribed circle when the size or location of the first maximum inscribed circle is different from the size or location of one of the second maximum inscribed circles; and

when the size or location of the first object is different from the size or location of one of the second object, adjusting the preset image template comprising the second object.

16. The object recognition method according to claim 15, further comprising:

comparing a size or location of the first modified object with sizes or locations of the second modified objects to determine whether to adjust the chamfer templates; and

when the size or location of the first modified object is different from the size or location of one of the second modified objects, adjusting the chamfer template comprising the second modified object.

17. The object recognition method according to claim 16, further comprising:

detecting a third centroid of the first modified object;

detecting a third maximum inscribed circle in the first modified object according to the third centroid;

detecting a fourth centroid of the second modified object;

detecting a fourth maximum inscribed circle in each of the second modified objects according to each of the fourth centroids;

determining whether a size or location of the third maximum inscribed circle is equal to a size or location of each of the fourth maximum inscribed circles; and

adjusting the chamfer template comprising the fourth maximum inscribed circle when the size or location of the third maximum inscribed circle is different from the size or location of one of the fourth maximum inscribed circles.

18. The object recognition method according to claim 15, wherein the chamfer distance transform is performed on edges or skeletons of the first object and the second objects.

19. The object recognition method according to claim 15, further comprising:

acquiring a provisional reference image comprising a third object;

performing the chamfer distance transform on the third object of the provisional reference image to produce a chamfer reference image comprising a third modified object;

determining whether difference between the first modified object and the third modified object is less than a second error threshold;

when the difference between the first modified object and the third modified object is less than the second error threshold, performing the step of determining whether the difference between the first modified object and one of the second modified objects is less than the first error threshold; and

replacing the provisional reference image by the real-time image when the difference between the first modified object and the second modified object is less than the first error threshold.

20. The object recognition method according to claim 19, wherein the chamfer distance transform is performed on an edge or a skeleton of the third object of the provisional reference image.

21. The object recognition method according to claim 15, wherein when movement of the first object in a sequence of next real-time images matches a preset trace model, the control command is sent, and the sequence of next real-time images is generated from a sequence of next original frames.

22. The object recognition method according to claim 21, wherein the first image capturing device has a field of view (FOV) comprising a preset sensing region, the preset sensing region is smaller than or equal to the FOV, and when the first object of the sequence of next real-time images moves out of the preset sensing region, the control command is sent to control the first image capturing device to perform image capturing.

23. The object recognition method according to claim 21, wherein the object recognition apparatus further comprises a second image capturing device, the first image capturing device has a first FOV to capture the real-time images, the second image capturing device has a second FOV, the first FOV is larger than or equal to the second FOV, and when the first object of the sequence of next original frames moves from the second FOV to the first FOV, the control command is sent to control the second image capturing device to perform image capturing.

24. The object recognition method according to claim 15, wherein the first object is a user'"'"'s hand.

25. An object recognition method, applied to an object recognition apparatus and comprising:

acquiring a real-time image, which comprises a first object, from a first image capturing device;

acquiring preset image templates each of which comprises a second object;

performing a chamfer distance transform on the first object of the real-time image to produce a chamfer image comprising a first modified object;

performing the chamfer distance transform on the second object of each of the preset image templates to produce a chamfer template comprising a second modified object;

determining whether difference between the first modified object and each of the second modified objects is less than a preset error threshold;

when the difference between the first modified object and one of the second modified objects is less than the preset error threshold, looking up a control command according to the preset image template corresponding to the second modified object;

when movement of the first object in a sequence of next real-time images acquired from the first image capturing device matches a preset trace model, sending the control command to control the object recognition apparatus;

comparing a size or location of the first object with sizes or locations of the second objects to determine whether to adjust the preset image templates, comprising;

detecting a first centroid of the first object;

detecting a first maximum inscribed circle according to the first centroid of the first object;

detecting a second centroid of each of the second objects;

detecting second maximum inscribed circles according to the second centroids of the second objects respectively;

determining whether a size or location of the first maximum inscribed circle is equal to a size or location of one of the second maximum inscribed circles; and

adjusting the preset image template comprising the second maximum inscribed circle when the size or location of the first maximum inscribed circle is different from the size or location of the second maximum inscribed circle; and

when the size or location of the first object is different from the size or location of one of the second object, adjusting the preset image template comprising the second object.

26. The object recognition method according to claim 25, wherein the first image capturing device has a FOV comprising a preset sensing region, the preset sensing region is smaller than or equal to the FOV, and when the first object of the sequence of next real-time images moves out of the preset sensing region, the control command is sent to control the first image capturing device to perform image capturing.

27. The object recognition method according to claim 25, wherein the object recognition apparatus further comprises a second image capturing device, the first image capturing device has a first FOV to provide the real-time images, the second image capturing device has a second FOV, the first FOV is larger than or equal to the second FOV, and when the first object of the sequence of next real-time images moves from the second FOV to the first FOV, the control command is sent to control the second image capturing device to perform image capturing.

28. The object recognition method according to claim 25, wherein the first object is a user'"'"'s hand.

29. The object recognition method according to claim 25, wherein the object recognition apparatus is a smart glasses.

30. An object recognition apparatus, comprising:

a first image capturing device, for recording a real-time image comprising a first object;

a storage device, for storing preset image templates each of which comprises a second object; and

a processing device connected to the first image capturing device and the storage device, for receiving the real-time image from the first image capturing device, receiving the preset image templates from the storage device, and performing the following steps;

performing a chamfer distance transform on the first object of the real-time image to produce a chamfer image comprising a first modified object;

respectively performing the chamfer distance transform on the second objects of the preset image templates to produce chamfer templates each of which comprises a second modified object;

determining whether difference between the first modified object and one of the second modified objects is less than a first preset error threshold;

when the difference between the first modified object and the second modified object is less than the first preset error threshold, looking up a control command according to the preset image template corresponding to the second modified object, the object recognition apparatus operating according to the control command;

comparing a size or location of the first modified object with sizes or locations of the second modified objects to determine whether to adjust the chamfer templates, and when the size or location of the first modified object is different from the size or location of one of the second modified objects, the processing device adjusting the chamfer template comprising the second modified object; and

detecting a third centroid of the first modified object, detecting a third maximum inscribed circle according to the third centroid of the first modified object, detecting a fourth centroid of each of the second modified objects, detects fourth maximum inscribed circles according to the fourth centroids of the second modified objects respectively, and determining whether a size or location of the third maximum inscribed circle is equal to a size or location of each of the fourth maximum inscribed circles; and

when the size or location of the third maximum inscribed circle is different from the size or location of one of the fourth maximum inscribed circles, the processing device adjusting the chamfer template comprising the fourth maximum inscribed circle.