CAPACITY AND ACCURACY ENHANCEMENT OF MOVING OBJECT TRACKING USING MULTIFEATURES
Moving object tracking can be start with ability to detect the object existence. System capabilities to understand visually recognize objects are still problem in research. Methods have been developed using location, moving direction, and velocity from previous frames as movement models. Moving obje...
Saved in:
| Main Author: | |
|---|---|
| Format: | Dissertations |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/40304 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Moving object tracking can be start with ability to detect the object existence. System capabilities to understand visually recognize objects are still problem in research. Methods have been developed using location, moving direction, and velocity from previous frames as movement models. Moving object tracking is still facing accuracy problems. Objects can be lost of detection because they change shapes and sizes due to distance changes, occlusion or hidden object, and illumination changes. Lost of accuracies in some methods can be balanced by other methods which still have accuracies. For capacity enhancement, some methods can be run sequentially interchangeable or in parallel concurrently to yield accuracies to follow object. Accuracies lost or decline also must be handled. Mathematical function approach is still rare in describing detail features for additional to object main features. Multiple Linear Regression in Choquet Integral can be used to describe pixel values as signal patterns which approximated by curve fitting methods. When moving, the pixel curve transforms with close value of coefficient of determinations, in addition to color filtering, and distance range restriction of previous position, but still traces similar object. This dissertation uses study cases of detecting detail attributes of moving human, some cases in smart meeting rooms, some cases in object tracking dataset with color attributes. Finally, comparison with existing methods shows that our proposed method works better in color tracking object in condition of object color difference from its surroundings, object broken movement in interval recorded videos, or fast motion object videos.
|
|---|