Object picking system, object detecting device, object detecting method

1. A device for detecting an object comprising a non-transitory computer readable medium with computer executable instructions stored thereon which, when executed by a processor, perform the following operations:

obtaining an image of objects from a visual sensor;

extracting one or more local characterized parts from the image;

detecting one or more object regions formed by a combination of the one or more extracted local characterized parts;

storing predetermined rotation angles of an object models corresponding to the objects, and object regions of the object model obtained for each predetermined rotation angle in a storage;

comparing the detected one or more object regions with the object regions of the object models stored in the storage unit, and extract one or more candidate object regions;

selecting a candidate object region as an operation target object region conforming to a predetermined determination reference among the one or more candidate object regions; and

outputting a rotation angle of the object model compared with the operation target object region as operation target object information,

wherein, when the one or more candidate object region is a single candidate object region, the single candidate object region is selected as the operation target object region, and

wherein, when the one or more candidate object region includes a plurality of candidate object regions, a candidate object region having a smallest absolute value of rotation angle among the plurality of candidate object regions is selected as the operation target object region.

2. The device of claim 1 , wherein the operations further comprise:

determining whether the detected one or more object regions are included within a predetermined error range by comparing the detected one or more object regions and the object regions stored in the storage; and

extracting the object region included in the predetermined error range as the one or more candidate object region.

3. The device of claim 1 , wherein the storage further stores the distance from the object model to the visual sensor, and wherein the object regions of the object model are obtained for each respective predetermined rotation angle and for each respective predetermined distance.

4. The device of claim 3 , wherein when the one or more candidate object regions include a plurality of candidate regions, a candidate object region is selected from among the plurality of candidate object regions compared with the object models, which has a rotation angle included in a predetermined range of the rotation angles of the object models, and has the smallest distance from the visual sensor, as the operation target object region.

5. The device of claim 1 , wherein the operations further comprise:

preprocessing the image obtained from the visual sensor before extracting the one or more local characterized parts.

6. An object picking system, the system comprising:

the device for detecting an object according to claim 1 ;

a picking robot controller; and

a picking robot accessed with the picking robot controller,

wherein the picking robot controller controls the picking robot to pick an operation target object corresponding to operation target object information.

7. A method of detecting an object, comprising:

obtaining an image of objects from a visual sensor;

extracting one or more local characterized parts from the image;

detecting one or more object regions formed by a combination of the one or more extracted local characterized parts;

comparing the detected one or more detected object regions with pre-stored object regions of object models corresponding to the objects, and extracting one or more candidate object regions;

selecting a candidate object region as an operation target object region conforming to a predetermined determination reference among the one or more candidate object regions; and

outputting a pre-stored rotation angle for the object model compared to the operation target object region as operation target object information,

wherein selecting the candidate object region includes, when the one or more candidate object regions is a single candidate object region, selecting the single candidate object region as the operation target object region, and

wherein selecting the candidate object region further includes, when the one or more candidate object regions include a plurality of candidate object regions, selecting the operation target object region based on pre-stored rotation angles for the object models compared with the plurality of candidate object regions and pre-stored distance from the visual sensor to the object model.

8. The method of claim 7 , wherein comparing the detected one or more object regions includes:

comparing the one or more detected object regions and the pre-stored object regions of the object models and determining whether the one or more detected object regions are included within a predetermined error range, and extracting the object region included in the predetermined error range as the one or more candidate object region.

9. The method of claim 7 , wherein the operation target object region corresponds to a candidate object region having a smallest absolute value of rotation angle among the plurality of candidate object regions.

10. The method of claim 7 , wherein selecting the candidate object region further includes:

when the one or more candidate object regions include the plurality of candidate object regions, selecting a candidate object region among the plurality of candidate object regions compared with the object models, which has a rotation angle included in a predetermined range of the rotation angles of the object models, and has the smallest distance from the visual sensor, as the operation target object region.

11. The method of claim 7 , wherein outputting the pre-stored rotation angle includes outputting the pre-stored rotation angle for the object model compared with the operation target object region and the pre-stored distance from the visual sensor as the operation target object information.

12. The method of claim 7 , further comprising:

pre-processing the obtained image,

wherein extracting one or more local characterized parts includes extracting the local characterized part from the pre-processed image.