Machine Vision - Pipe Recognition

Title: Machine Vision - Pipe Recognition

​

Date: 2018.10 - 2018.12

​

Class Type: Special Topics in Architecture and Urban Design

​

Instructor: Kaiho Yu, Marta Nowak

​

Team member: Tingji Guo, Yanxu Jia, Qinxuan li

​

Introduction:

The aim of this project was to introduce a theory, application, and techniques of machine vision in conjunction with image production in an architectural context. My team created an understanding of the problems involved in machine vision and object recognition in architectural scenarios, such as pipe detection. We used drone as a tool, by using algorithms for image processing and pattern recognition. From this analysis, we set up workflows of classification, localization, object detection, and instance segmentation for robots within architecture or even urban scale.