Method of searching data to identify images of an object captured by a camera system
A camera system comprises an image capturing device, object detection module, object tracking module, and match classifier. The object detection module receives image data and detects objects appearing in one or more of the images. The object tracking module temporally associates instances of detected objects, each of which has a signature representing features of the detected object. The match classifier matches object instances by analyzing data derived from the signatures. The match classifier determines whether the signatures match.
1. A system comprising:
an object detection module operable to receive image data comprising one or more images captured by an image capturing device and detect an object that appears in the one or more images;
an object classification module operable to determine a classification label for the object based on the one or more images; and
a video analytics module operable to receive metadata comprising the classification label and appearance characteristics derived from the one or more images, generate a distance measure based on the metadata, compare the distance measure to a threshold, determine based on the comparison that the object matches a reference object, and send a notification indicating a match between the object and the reference object, wherein the object and the reference object are instances of a same physical object.
2. The system of claim 1 , further comprising:
an object search module operable to eliminate one or more candidate objects from being considered as matches for the object based on detection of the candidate objects prior to a predetermined time frame.
3. The system of claim 1 , further comprising:
an object search module operable to eliminate one or more candidate objects from being considered as matches for the object based on additional classification labels previously determined by the object classification module for the candidate objects.
4. The system of claim 1 , further comprising:
a metadata database operable to store the metadata and additional metadata for candidate objects detected by the object detection module and provide the additional metadata upon request;
an object search module operable to retrieve the additional metadata from the metadata database and, based on the additional metadata and in response to a user command received via an input device, identify a group of the candidate objects that match the object, wherein the candidate objects in the group are instances of the same physical object; and
a user interface connected to the object search module, the user interface being operable to select the object based on the user command received via the input device and present representations of the candidate objects in the group identified by the object search module on a display in response to the user command.
5. The system of claim 4 , wherein the object is a human object, the representations of the candidate objects in the group are video clips, and the additional metadata for the candidate objects are linked in the metadata database to locations where the candidate objects were detected within fields of view of additional image capturing devices.
6. A method comprising:
receiving image data comprising one or more images captured by an image capturing device;
detecting an object that appears in the one or more images;
determining a classification label for the object based on the one or more images;
receiving metadata comprising the classification label and appearance characteristics derived from the one or more images;
generating a distance measure based on the metadata;
comparing the distance measure to a threshold;
determining based on the comparison that the object matches a reference object; and
sending a notification indicating a match between the object and the reference object, wherein the object and the reference object are instances of a same physical object.
7. The method of claim 6 , further comprising:
eliminating one or more candidate objects from being considered as matches for the object based on detection of the candidate objects prior to a predetermined time frame.
8. The method of claim 6 , further comprising:
eliminating one or more candidate objects from being considered as matches for the object based on additional classification labels previously determined for the candidate objects.
9. The method of claim 6 , further comprising:
receiving a user command via an input device;
retrieving additional metadata from a metadata database for previously detected candidate objects;
in response to the user command and based on the additional metadata, identifying a group of the candidate objects that match the object, wherein the candidate objects in the group are instances of the same physical object; and
presenting representations of the candidate objects in the group on a display in response to the user command.
10. The method of claim 9 , wherein the object is a human object, the representations of the candidate objects in the group are video clips, and the additional metadata for the candidate objects are linked in the metadata database to locations where the candidate objects were detected within fields of view of additional image capturing devices.