Image processing device and image processing method
An image processing device includes a detection unit, a feature amount generating unit, a tracking unit, a feature amount change detecting unit, a tracking history extracting unit, and a determination unit, so that when occlusion in a selected cluster has occurred, the image processing device can determine whether or not objects are identical without using data of an image in which the occlusion has occurred. An image in which the occlusion has occurred can be excluded by using the data of the image determined to have no change in the feature amount, whereby a decrease in accuracy of determining whether or not the objects are identical can be suppressed.
1 . An image processing device comprising:
processing circuitry configured to
analyze a first image sequence including a plurality of images obtained by imaging a first region to detect a first object that appears in any one of the images of the first image sequence, and analyze a second image sequence including a plurality of images obtained by imaging a second region different from the first region to detect a second object that appears in any one of the images of the second image sequence;
generate a first object feature amount for tracking the first object and a second object feature amount for tracking the second object;
track the first object in a time direction within the first image sequence using the first object feature amount to output a tracking result including a plurality of values of the first object feature amount as a first object tracking history, and track the second object in the time direction within the second image sequence using the second object feature amount to output a tracking result including a plurality of values of the second object feature amount as a second object tracking history;
detect whether or not the first object feature amount has changed using the first object tracking history to output a detection result as a first feature-amount-change detection result, and detect whether or not the second object feature amount has changed using the second object tracking history to output a detection result as a second feature-amount-change detection result;
extract a value of the first object feature amount from the first object tracking history for an image determined to have no change in the first object feature amount to output the extracted value of the first object feature amount and the first object tracking history as a first tracking-history extraction result, and extract a value of the second object feature amount from the second object tracking history for an image determined to have no change in the second object feature amount to output the extracted value of the second object feature amount and the second object tracking history as a second tracking-history extraction result; and
determine whether or not the first object and the second object are identical using the first tracking-history extraction result and the second tracking-history extraction result.
2 . The image processing device according to claim 1 ,
wherein the processing circuitry is further configured to
estimate an imaging surface of the first object from the first object tracking history to output the estimated imaging surface as a first imaging surface, and estimate an imaging surface of the second object from the second object tracking history to output the estimated imaging surface as a second imaging surface;
extract a first comparison feature amount for the first imaging surface of the first object, and extract a second comparison feature amount for the second imaging surface of the second object; and
determine whether or not the first object and the second object are identical using the first comparison feature amount and the second comparison feature amount, when the first imaging surface and the second imaging surface are the same.
3 . The image processing device according to claim 2 , wherein
the first object tracking history includes a prediction value of the first object feature amount at a time of interest, the prediction value being predicted on a basis of an observation value of the first object feature amount observed at the time of interest and an observation value of the first object feature amount observed at a past time preceding the time of interest, and
the processing circuitry detects whether or not the first object feature amount has changed on a basis of the observation value and the prediction value of the first object feature amount at the time of interest.
4 . The image processing device according to claim 3 , wherein
the first object tracking history includes a speed of the first object, and
the processing circuitry estimates the first imaging surface from the speed of the first object.
5 . The image processing device according to claim 2 , wherein
the processing circuitry generates, on a basis of an image feature amount obtained from data of a plurality of images corresponding to a set of the first objects having a same imaging surface, the first comparison feature amount for the set.
6 . An image processing method performed by an image processing device, the method comprising:
analyzing a first image sequence including a plurality of images obtained by imaging a first region to detect a first object that appears in any one of the images of the first image sequence, and analyzing a second image sequence including a plurality of images obtained by imaging a second region different from the first region to detect a second object that appears in any one of the images of the second image sequence;
generating a first object feature amount for tracking the first object and a second object feature amount for tracking the second object;
tracking the first object in a time direction within the first image sequence using the first object feature amount to output a tracking result including a plurality of values of the first object feature amount as a first object tracking history, and tracking the second object in the time direction within the second image sequence using the second object feature amount to output a tracking result including a plurality of values of the second object feature amount as a second object tracking history;
detecting whether or not the first object feature amount has changed using the first object tracking history to output a detection result as a first feature-amount-change detection result, and detecting whether or not the second object feature amount has changed using the second object tracking history to output a detection result as a second feature-amount-change detection result;
extracting a value of the first object feature amount from the first object tracking history for an image determined to have no change in the first object feature amount to output the extracted value of the first object feature amount and the first object tracking history as a first tracking-history extraction result, and extracting a value of the second object feature amount from the second object tracking history for an image determined to have no change in the second object feature amount to output the extracted value of the second object feature amount and the second object tracking history as a second tracking-history extraction result; and
determining whether or not the first object and the second object are identical using the first tracking-history extraction result and the second tracking-history extraction result.