透過雙目立體視覺實現四輪移動自主機器人之避障控制

Abstract

本文以四輪自主移動機器人作為控制系統搭配雙目立體視覺進行環境資料的收集，以實現環境適應導航與避障功能，本文主要分成三個部份，包含馬達控制部分、影像測距區塊以及避障處理區塊，透過微控制器(microcontroller)作為馬達控制的控制核心，筆記型電腦(Notebook Computer)與兩個網路攝影機(Webcam)進行前方障礙距離的計算以及避障控制。透過霍爾元件偵測兩輪的車輪轉速，並將轉速資料回傳給微控制器用來完成速度追蹤與避障控制。為了透過視差測距完成環境檢測，必須要找出兩個攝影機中的同一個空間座標位置，本文使用正規化相關匹配法演算法(Normalized Cross Correlation method, NCC)進行匹配點搜尋以獲得視差值，再利用相似三角形定理計算相對於四輪自主移動機器人的環境距離資料，最後透過模糊控制(Fuzzy Control)來實現避障功能。This thesis develops a four-wheeled autonomous mobile robot with binocular stereo vision that collects distance data and uses the fuzzy control to achieve obstacle avoidance. The four-wheeled autonomous mobile robot possesses the functions of speed control, image ranging and obstacle avoidance. The speed control uses a microcontroller to control the left and right wheels. Hall sensors are used to detect the wheel speed of two wheels, and the microcontroller receives the data of the Hall sensors and then calculates the real speed. The function of image ranging is implemented by two image sensors. The fuzzy logic is used to achieve the obstacle avoidance function. In order to complete the function of parallax ranging in environmental detection, it is necessary to find out the same spatial coordinate position in two image sensors. In this thesis, normalized cross correlation method (NCC) algorithm is used to perform point matching search to calculate parallax, and then similar triangle theorem is used to calculate the environmental distance data relative to the autonomous mobile robot.