Track & Assign Unique IDs to Objects in Video c#

问题:

I am new to Computer Vision and i am using EmguCV library. My goal is to track and assign unique IDs to all objects in a video and i'm almost lost!!.

Approach 1: for each frame in the video:

• Background subtraction using (BackgroundSubtractorMOG)
• Get Blobs using (CvBlobDetector)
• Track detected Blobs using (CvTracks)

The issues in this approach are:

• The same object (ex: person) may be detected as more than one Blob and each one is assigned a different ID.
• If the person/object left the scene or gets occluded by another object in one frame, it is assigned a new ID (lose tracking).

Approach 2: for each frame in the video:

• Detect objects in the frame using YOLO
• Use modern trackers in EmguCV (TLD, KCF) Trackers

The issues in this approach are:

• How to map detected objects in YOLO to trackers in EmguCV?
• If the person/object left the scene or gets occluded by another object in one frame, i lose tracking.
• How to identify the same object in different frames (using color/appearance/features not the location)? what is the best method?

Any help/ideas for these approaches will be very appreciated or new approaches for tracking and ID tagging.