|
Hybrid Deep Learning and Efficient Segmentation Technique for Multiple Object Tracking
|
|
| Author:
|
S.SATHYANARAYANAN, DR.ANNA SARO VIJENDRAN
|
| Abstract:
|
Tracing many objects in images or videos is a computer vision problem. This issue has been gaining importance both commercially and academically. The technique called MOT (Multi-Object Tracking) is localizing object movements within a specific period of time. Tracing trajectories of objects in movements and analyzing them post their own set of issues. This research work proposes a hybrid technique involving CNN (Convolution Neutral Networks) and SVM (Support Vector Machine) for tracing multiple objects within frames. The proposed work is a set of simple algorithmic techniques whose performances in simulations are accurate. Further, the proposed work also uses other techniques in its segmentation and pre-processing of images or frames.
|
| Keyword:
|
DBCWMF (Decision-based Coupled Window Median Filter), DWT (Discrete Wavelet Transform), PSO (Particle Swarm Optimization), SIFT (Scale Invariant Feature Transform), CRF (Conditional Random Field), KLT (Kanade-Lucas-Tomasi), CNN, SVM.
|
| EOI:
|
-
|
| DOI:
|
https://doi.org/10.31838/ijpr/2020.12.04.546
|
| Download:
|
Request For Article
|
|
|