Scholarlink

Since unmanned aerial vehicle (UAV) has gained popularity in the current era, UAV has been used as a form of surveillance. UAVs are known for its portable, easy to control, and unmanned features, which makes it the finest choice to install a surveillance system on. With the high rates of the property occurring in Malaysia, the idea of creating a newly improved and state-of-the-art CCTV pose as a solution to combat the stated issue. The obtained system integrates both human detection and tracking algorithm as well as face recognition system to detect trespassers caught on footage and to recognise a person’s facial attributes. The underlying motivation for the project had emerged due to the high number of cases for property crimes occurring locally and uprising surveillance applications of UAV. Upon implementing the algorithm, the obtained system must be able to fulfil the objectives, such as to critically investigate the potential of using UAV for human tracking and detection in lowering property crime cases, modify an algorithm for detecting or classifying human with an improvement of 1.14% from aerial images captured by a UAV, and apply face recognition system within the acceptable range of distance on aerial images captured. The project research would be split into 3 stages, which are code implementation for human tracking and detection, code implementation for facial recognition, and improvement of code implemented in the previous stages. During these stages, the results of the existing algorithms would be recorded on a spreadsheet software and issues emerging from the algorithm would be aimed to be resolved. In stage 1, the tested algorithm is expected to run with an improvement of 1.14%. In stage 2, testing with the algorithm built to output the acceptable distance range for the project. In stage 3, the combined algorithm is expected to work well with the footage taken in CC lab at Taylor’s Block E. The expected outcome of the proposed system is to be able to detect and track people in the CC lab, to classify a human face, to prove that UAV utilisation may have a potential in lowering cases of property crimes. Tiny YOLOv3 and Haar Feature-based Cascade Classifiers has been utilised for human detection and tracking and face recognition algorithm respectively and improvements on each stage are still in the works.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning-Aided Aerial Surveillance: Enhancing Trespasser Detection Using UAVs

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Swee King Phang,
  • Farah Hani Ma’amor,
  • Afizan Azman

摘要

Since unmanned aerial vehicle (UAV) has gained popularity in the current era, UAV has been used as a form of surveillance. UAVs are known for its portable, easy to control, and unmanned features, which makes it the finest choice to install a surveillance system on. With the high rates of the property occurring in Malaysia, the idea of creating a newly improved and state-of-the-art CCTV pose as a solution to combat the stated issue. The obtained system integrates both human detection and tracking algorithm as well as face recognition system to detect trespassers caught on footage and to recognise a person’s facial attributes. The underlying motivation for the project had emerged due to the high number of cases for property crimes occurring locally and uprising surveillance applications of UAV. Upon implementing the algorithm, the obtained system must be able to fulfil the objectives, such as to critically investigate the potential of using UAV for human tracking and detection in lowering property crime cases, modify an algorithm for detecting or classifying human with an improvement of 1.14% from aerial images captured by a UAV, and apply face recognition system within the acceptable range of distance on aerial images captured. The project research would be split into 3 stages, which are code implementation for human tracking and detection, code implementation for facial recognition, and improvement of code implemented in the previous stages. During these stages, the results of the existing algorithms would be recorded on a spreadsheet software and issues emerging from the algorithm would be aimed to be resolved. In stage 1, the tested algorithm is expected to run with an improvement of 1.14%. In stage 2, testing with the algorithm built to output the acceptable distance range for the project. In stage 3, the combined algorithm is expected to work well with the footage taken in CC lab at Taylor’s Block E. The expected outcome of the proposed system is to be able to detect and track people in the CC lab, to classify a human face, to prove that UAV utilisation may have a potential in lowering cases of property crimes. Tiny YOLOv3 and Haar Feature-based Cascade Classifiers has been utilised for human detection and tracking and face recognition algorithm respectively and improvements on each stage are still in the works.