3D deep learning-based sensor placement optimization for personalized ageing-in-place

This report presents a methodology for optima sensor placement in indoor setup using PointNet++ for 3D semantic segmentation followed by a coverage-based algorithm. The process begins with the pre-processing of point cloud data, followed by semantic segmentation using PointNet++ to identify key obje...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Wei, Yao
مؤلفون آخرون: Shen Zhiqi
التنسيق: Final Year Project
اللغة:English
منشور في: Nanyang Technological University 2023
الموضوعات:
الوصول للمادة أونلاين:https://hdl.handle.net/10356/166039
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
المؤسسة: Nanyang Technological University
اللغة: English
الوصف
الملخص:This report presents a methodology for optima sensor placement in indoor setup using PointNet++ for 3D semantic segmentation followed by a coverage-based algorithm. The process begins with the pre-processing of point cloud data, followed by semantic segmentation using PointNet++ to identify key objects in the scene, achieving point mIoU OF 54.9\% and voxel mIoU of 54.4\% on the evaluation set. To enhance the accuracy, a post-processing step with DBSCAN is implemented. The sensor placement algorithm then calculates the coverage for each candidiate sensor location, taking into consideration the sensing range, angle, and visibility through obstacles. The method is tested on various real-life environments, including a medium-sized bedroom and a large living room. Results demonstrate the efficacy of the approach, with sensor coverage ranging from 87\% to 99\% in the bedroom and up to 65\% for a single sensor in the living room. However, limitations include the time complexity of the algorithm and its performance on more complex floor plans. The study provides a foundation for further development in optimizing sensor placement in indoor environments while considering factors such as connectivity and interaction between multiple sensors.