本論文提出一個可以增加速度及可靠度的高速自動對焦策略,本策略乃是整合灰色預測理論及模糊推論理論之高速對焦演算法則,首先將獲得的影像經過清晰尺度演算法計算後得到全域幅度變化量及區域幅度變化量,將這二個變化量當作是模糊控制器的輸入變數,經模糊控制器運算後得到輸出步距。當全域幅度變化量或區域幅度變化量減少到非正值時就滿足灰色預測啟動條件,若預測出來的預測值為對焦曲線的峰值後,將鏡頭返回到對焦曲線的最佳點。此策略不但可以減少對焦所花費的時間,更可以減少馬達在最佳點附近移動的次數,最後本論文所提的演算法實現於安裝有Microsoft Windows和即時作業子系統﹝RTX﹞的PC來驗證演算法的效能。本論文所提的演算法與傳統的二分法來做比較,由實驗結果看出本論文所提的演算法可以減少對焦花費的時間。
This thesis proposes a high speed auto-focusing strategy which dramatic increasing the speed and improving the reliability of the auto-focusing technique. This strategy integrates the fuzzy reasoning and grey prediction algorithm. Firstly, the local and global slopes of sharpness function, calculated by the specified image caught by CCD, are feeding into fuzzy reasoning scheme as input variables. The corresponding moving step is calculated from fuzzy reasoning scheme. Then, the gray prediction model is adapted to predict the peak of the sharpness function curve after the local or global slopes decreasing. Therefore, the focusing mechanism comes back to the previous position which is the focusing position. The strategy can reduce focusing time around the focusing position. Finally, an experimental setup, implemented on a PC with Microsoft windows and RTX subsystem, is installed to verify the performance of proposed strategy. Comparing the experimental results of proposed strategy with traditional binary-search algorithm, the results reveal that this strategy can reduce the focusing time.