基於視覺伺服與離散時間預估法的無人載具控制

時至今日，基於視覺的無人飛行載具控制已經是相當熱門的題目。因為科技的進步，可以實現即時的影像處理。基本上這個控制問題可以分成兩個部分，目標位置的估測以及無人飛行載具的控制系統設計。如何應付多變的自然環境是設計控制系統的一大考驗。在許多不同的方法中，本論文選用了結合視覺伺服與離散時間預估法來達成控制無人飛行載具追蹤地面目標的行為。估測部分使用卡爾曼濾波器以及連續適應性均值飄移演算法提高目標位置的估測準確率。視覺伺服藉由現在位置與理想位置的距離創造一個最佳解給離散時間預估法，由離散時間預估法控制無人飛行載具的動作。無人飛行載具會不斷地調整自身姿態，直到抵達視覺伺服控制計算出來的最佳解。

關鍵字

卡爾曼濾波器；連續適應性均值飄移演算法；視覺伺服；離散時間預估法

並列摘要

On these days, unmanned aerial vehicle (UAV) target tracking based on vision is a popular research topic. Because technology has improvement, that a real time image processing can be realized. Basically, it can separate to two main parts, estimate the target position first, then design a controller system of the UAV motion. And the coordination of two parts will influence the precision of result. One of the difficult challenges of UAV tracking target is that control system must be stable enough to face unpredictable natural airstream. There have many different methods to reach the purpose. This thesis presents a solution that combines a constrained visual servo and one-step ahead prediction control to achieve tracking ground target by UAV. Kalman filter and Camshift are used in tracking algorithm, which increase the accuracy of target object’s position. Constrained visual servo controller calculates the distance between desired position and object position which creating an optimize solution to DOPC controller. And then DOPC controller based on the result to decide the UAV motions. The UAV will constantly adjust its posture, until reaching the optimize position by constrained visual servo controller calculation.