裴姍 章騰
摘 要:目前關于幾何圖形識別的算法存在的問題主要有準確率低,計算復雜度較高,運行時間較長等。利用簡潔高效的Freeman鏈碼算法結合幾何圖形特有的幾何屬性設計出新的算法,使其能夠快速識別幾何圖形的頂點分布,并反映在一張邊界震蕩(boundary vibration,BV)曲線圖上。該曲線圖能夠刻畫幾何圖形的屬性,如頂點的個數、距離分布,角度的大小,線段的曲直和圖形的周長等。因此通過對曲線圖的特征分析可以準確識別對應的幾何圖形。該算法不受圖形的平移、旋轉、放縮、噪聲影響。為了測試算法的穩定性,仿真試驗針對九種不同的隨機生成且帶噪聲的幾何圖形,結果識別率較高,運行速度較快,達到了預期的效果。
關鍵詞:幾何圖形;識別;Freeman鏈碼;邊界震蕩曲線
中圖分類號:TP317.4 文獻標識碼:A
Abstract: Currently there exist some problems about recognition algorithm for geometry figure,such as low rate of accuracy,high order of complexity,long time of running and so forth.This article designs a new algorithm by using the laconic efficient Freeman chain code algorithm with geometry properties of geometry figures,so that it can recognize the distribution of vertexes of geometry figures quickly,and it is displayed in a figure of boundary vibration(BV) curve.By analyzing,we see this figure of curve can describe the properties of geometry figures,such as the number and distance distribution of vertexes,the size of angles,the curvity of line segments,the perimeter of figures and so on.Hence we can recognize the relevant geometry figure accurately by analyzing the characteristic of the figure of curve.This algorithm isn't influenced by translation,rotation,extension and noise of figures.In order to test the stability of the algorithm,the simulation is in connection with nine different kinds of geometry figures which are generated arbitrarily and carry noise,the result shows the rate of accuracy is high and the speed of running is fast,we acquire the prospective effect.
Key words: geometry figure;recognition;Freeman chain code;boundary vibration curve
目前圖像識別技術應用十分廣泛,如在實驗室監控中的應用[1],在編組站駝峰作業過程控制中的應用等[2]。也有學者專門研究了圖像識別的技術現狀與發展趨勢[3]。而對基本的幾何圖形的識別在實際的圖像識別中十分關鍵,通過幾何圖形的識別可以快速掌握實際圖像的輪廓,屬性等特征。目前已有許多學者在幾何圖形的識別算法研究中卓有成效,文獻[4]中依據多邊形頂點和其它邊緣像素點的特征值變化規律設計了一種簡潔高效的識別算法;文獻[5]中采用了霍夫變換理論來研究圖形的識別。另外,還有一些學者的研究基于深度學習、神經網絡、遺傳算法、特征分布等技術與方法[6-9]。……