利用多個超音波感測器與卡曼濾波器實現單一物體之追蹤

Abstract

本論文之目的為利用超音波測距原理，在長59公分，寬32.4公分之大型面板 上，定位與追蹤物體所在的二維座標。這個系統包含了裝置在螢幕邊框的一個超音波發射器與五個接收器，藉由每個通道之時差測距，計算出物體的座標位置。超音波在傳送過程中的衰減，以及不同目標物形狀、大小與相對於感測器之方向，皆會對超音波回聲產生干擾並造成波形之變化，進而影響到時差測距的準確度。本論文針對以上之問題提出一個新的時差測距之估測方法，可尋找出合理的反射波區間，再藉由反射波包絡之雙指數模型與牛頓最佳化演算法，得到更精準與穩定的時差測距。本論文亦使用擴展型卡曼濾波器來估測物體座標位置，可同時考慮時差測距之干擾與離群值的問題對於座標位置之影響，有效地在雜訊環境中提升物體定位與追蹤之穩定度。最後藉由不同情境的實驗來測試本系統之表現，包含了以麥克筆與人類手指作為靜止目標物之位置估測，與移動的人類手指之追蹤。

This thesis realizes 2D coordinate localization and tracking of single object on a 59-centimeters-long and 32.4-centimeters-wide large panel based on the principle of ultrasonic range measurement. There are 1 ultrasonic transmitter and 5 receivers equipped on the edge of the panel. The time-of-flight (TOF) from each channel is first derived and the target coordinate is then obtained. TOF estimation is inaccurate due to shape distortion of echoes waveform, which is caused by attenuation during propagation and also varies with target type, size, location, and orientation. A new method of TOF is presented to solve the above problem. This method provides a more accurate and steady TOF estimation by fitting the double exponential model on the reasonable region of envelope using Newton-Ralphson optimization. An Extended Kalman Filter is also designed to estimate the target coordinate inherently accounting the interference and outlier issue of the derived TOF, and effectively reduces the disturbance to target localization in critical measurement condition. The system performance was assessed through experimental evaluation of several scenarios, including localization of stationary marker pen, stationary human’s finger, and tracking on moving human’s finger.