Jul 21, 2021 | Vanshika Kaushik
Object tracking is used for surveillance, it is also used for medical imaging, traffic flow monitoring and human computer interaction. Some tasks in object tracking are difficult to be performed by computers. Recognition of objects of static objects in video footage is still a challenge in this field. It is difficult to achieve optimum level of accuracy in multi object tracking.
Fast R- CNN, Region based convolutional neural networks, single shot detector (SSD) are used for object tracking. Researchers at Gwangju Institute of Technology, Korea have developed a multi- object tracking technique that employs deep learning for object tracking.
(Must Check: Streaming Android Camera using OpenCV)
Technique (Deep- TAMA) will provide new solutions in the object tracking field. This tracking model will extract some known features of detected objects. Objects that are detected by the technique will be compared to other objects which share features with detected objects. Researchers integrated joint interface neural networks with long short term memory networks (LTSM).
LTSM will connect its stored appearances with previously detected object appearances. Based on the previous data algorithm will look into visual occlusions. This method will provide accurate detection in a variety of public surveillance tasks. Multi object tracking algorithm can be used for maximizing accuracy in autonomous driving.
According to Tech Explore, Professor Moongu Jeon said, “We believe our methods can inspire other researchers to develop novel deep-learning-based approaches to ultimately improve public safety."