Thermal cameras have long been acknowledged to have an advantage over visible-light cameras, specifically in low-light conditions, such as at night. However, the biggest challenge in using one is when occlusion (when the object is covered) and thermal crossover (when the object has a similar thermal appearance) happen.
Because of that, Dr. Nur Ibrahim, a Doctor from the Faculty of Engineering (FT) of Universitas Indonesia (UI), developed a new method to increase the effectiveness of nighttime surveillance systems by utilizing thermal cameras. He made an innovation in the development of the complex negative example method which can reduce the possibility of false detection and identity switching.
A complex negative example is data that has no objects in it, or only a small part of the objects in the data. This will help the system distinguish observed objects when these objects are close to or covered by other similar objects around them.
“Research data was taken using a compact thermal camera installed on an Android device. Data retrieval was done at night with the occlusion and thermal crossover scenarios in the forest, where the objects observed were humans moving across the camera at various distances and weather conditions,” said Dr. Nur.
The result shows that the developed method has a significant impact on increasing the surveillance system performance. In the visual tracking system, the addition of the interval type 2 fuzzy logic system (IT2 FLS) algorithm to weighted multiple instance learning (WMIL) succeeded in increasing the success rate by 10% to 14% and precision by 0.21 to 0.33.
This research, done by Dr. Nur, is a significant contribution to the object-tracking field, especially for nighttime surveillance applications. The optimization of the WMIL method has succeeded in increasing success rate and precision, while the performance improvement of you-only-look-once (YOLO) and deep appearance-based trackers reduce the potential of false detection and identity switching significantly.
The Dean of FTUI, Prof. Ir. Heri Hermansyah, S. T., M.Eng., IPU, stated, “With this innovation, surveillance systems are hoped to be more effective in monitoring objects movements in the nighttime, improve security, and give a faster response to incidents. This research paved the way for further development in operating thermal cameras for various security and monitoring applications in the future.”
Owing to his research, titled, “Pengembangan Metode Complex Negative Example dalam Sistem Object Tracking Berbasis Deep Appearance Features pada Citra Thermal,” Dr. Nur managed to attain a Doctoral degree in the Electrical Engineering field in the FTUI Doctoral promotion open session on Friday, June 28. This research has also been published in multiple journals and leading international conferences, such as the Journal of Images and Graphics (SCOPUS Q2) and IEEE 3rd International Conference on Robotics Automation and Artificial Intelligence (RAAI) 2023, as well as submitted to IJTECH (SCOPUS Q1).
Dr. Nur Ibrahim successfully defended his dissertation and attained the Cum Laude predicate with a GPA of 3.90. He graduated as the 171st doctor of the Electrical Engineering Study Program and the 548th in FTUI. The promotion session was chaired by Prof. Ir. Mahmud Sudibandriyo, M.Sc., Ph.D, with Prof. Dr.Eng. Drs. Benyamin Kusumoputro, M.Eng. as promoter, and Dr. Ir. Aries Subiantoro, M.Sc. as co-promoter. As for examiner team, it consists of Prof. Dr. Ir. Feri Yusivar, M.Eng.; Dr. Abdul Halim, M.Eng.; Dr. Abdul Muis, S.T., M.Eng.; Prof. Dr. Ir. Mauridhi Hery Purnomo, M.Eng.; and Dr. Muhammad Rif’an, S.T., M.T.
