左右暨前後視覺導航法小船實驗探討

A Small Boat-Based Experimental Studies on the Left-Right and Front-Rear Leading Mark Visual Guidance Methods

本文主旨在以FRP實驗船舶為測試平台，探討以影像處理技術，利用前後疊標、左右疊標、左右加後助航疊標及修正式左右加後助航疊標共四種疊標導航法，並結合內模式與模糊控制器，完成船舶自動導引進港試驗之目的。實驗之進行乃透過架設於實驗船船艏之CCD攝影機來擷取岸壁上預設之疊標影像，利用LABVIEW/Vision Builder AI軟體系統架構進行色彩空間轉換，經由適當選取影像成分中H（Hue，色調）、S（Saturation，飽和度）與V（Value，色深度）篩選值域之最大及最小值，達到疊標鎖定之目的，並計算疊標中心與CCD攝影機中心之相對角度關係，送至自航控制器產生舵角輸出，用以導引船舶至預定航道內。最後於國立台灣海洋大學小艇碼頭水域進行相關之實船測試，以確認本文所提方案之可行性，並進行各類導航法與控制器的性能結果比較及分析。實驗結果顯示，前後疊標導航法結合模糊控制器整體之軌跡收斂行為優於結合內模式控制器，左右疊標導航法則呈現近似單點導航之效果。而左右加後助航疊標導航法其整體導航軌跡收斂效果與前後疊標導航法相似，而最後一型修正式左右加後助航疊標導航法結合內模式控制器確實有加速收斂之效果，但若結合模糊控制器時若控制器之設計不理想時，其輸入角度意義上將與左右疊標導航法雷同，將造成船舶軌跡收斂上呈現單點導航之趨勢。 This study uses a small FRP boat as the experimental test platform to examine the performance of four types of image processing-based visual guidance schemes to guide a ship towards a prescribed straight water channel. Specifically, the four types of visual guidance methods are the fore-aft leading mark method, the left-right leading mark method, the left-right with additional rear leading mark method and the modified left-right with additional rear leading mark method. The LABVIEW/Vision Builder AI system working environment is adopted for the design of the Fuzzy Controller and the Internal Model Controller (IMC). A CCD camera mounted on the bow of the FRP boat is used to acquire images of the leading marks. The color recognition scheme and noise filtering are adopted in the image processing and proper threshold values are selected to ensure correct target acquirement. The geometric centers of the two leading marks are then calculated, and the separation distance between them is used to compute the deviated heading angle needed as feedback error signal in the adopted autopilots. The feasibility of the proposed concept was verified by performing a series of FPR boat experiments conducted in the waterways of the NTOU small boat harbor. Experimental results indicate that the fore-aft leading mark guidance method with Fuzzy controller has better performance than with the IMC controller, the left-right leading mark guidance scheme is less satisfactory because of its single leading mark like performance, and the performance of the left-right with additional rear leading mark method is similar to the fore-aft leading mark method. Finally, the modified left-right with additional rear leading mark method combined with the IMC controller has better trajectory convergence rate. However, the modified left-right with additional rear leading mark method combined with fuzzy controller might exhibit single leading mark guidance behavior, if the fuzzy input variables are not properly defined.