應用ImageJ於估算綠美化工程之植草存活率

Using ImageJ for Estimating the Survival Rate of Green Grass

張逸群(Yi-Chun Chang)

指導教授 ： 鍾 文 貴

國立屏東科技大學/工學院/土木工程系所/碩士(2013年)

https://doi.org/10.6346/NPUST.2013.00044

摘要

綠美化工程為各項環境工程應用與改造不可或缺的項目，其中植草是工程執行中之基本元素。本文主要的目的在於使用一般市售數位相機與免費影像處理軟體ImageJ，建立一套簡易的植草存活率估算分析法，作為綠美化工程品質管理與驗收之參考。

本研究主要以映射投影法、顏色轉換法與門檻閥值法等三種影像處理方法，針對草地影像進行影像處理。由於影像分割是影像辨識的基礎，而顏色模型與門檻閥值方法的選擇常是影響影像處理分析的關鍵。因此，本研究在建立一套可靠與客觀的影像處理分析系統前，必須尋找較為合適的顏色模型與門檻閥值法組合，以尋求最佳的影像處理分析結果。

為了從影像處理軟體ImageJ所提供的顏色模型與門檻閥值法中，客觀的評價與篩選出最佳影像分割方法，本研究透過以下的評價方式來進行篩選:

(1)透過在室內不同拍攝條件下，統計分析裸土影像灰階值之標準差；依標準差的大小，評價顏色模型分量對於不同品質的裸土影像，其辨識裸土顏色特徵的可靠度。

(2)透過在室內固定拍攝條件下，統計分析植草影像灰階值與植草生長天數的相關係數；依相關係數的大小，評價顏色模型分量對於辨識植草生長影像的綠度變化，其辨識草坪綠顏色特徵的敏感度。

(3)最後以影像處理軟體ImageJ分析計算的植草存活率結果，與點框法調查計算值作比較，依據絕對誤差、平均絕對誤差與影像分割質量等，評估影像處理分析系統和影像分割方法的可行性與準確性。

經過客觀的統計評價與篩選後，發現在顏色模型Lab中，顏色模型分量a具有最低的標準差與最高的相關係數，顯示顏色模型分量a在辨識草地顏色特徵上具有一定的可靠度與敏感度；用顏色模型分量a與門檻閥值法Yen的組合，進行戶外草地存活率的估算工作，絕對誤差介於0.057% ~ 6.052%，而平均絕對誤差(MAPE)小於5%，顯示本研究所建立的影像處理方法具備一定程度的準確性，在實務上具有實用價值。

關鍵字

ImageJ ； 植草存活率 ； 顏色模型 ； 門檻閥值法

並列摘要

The green landscaping works for the environmental engineering applications, and the transformation of an integral part of the project, which grass is the basic element in the project execution. The main purpose of this paper is to use generally commercially available digital camera and image analysis software ImageJ establishing a simple analysis method for estimating the grass survival rate, as the greening project quality management and acceptance of reference.

In this research, we use the Image segmentation, the Color transform and the Threshold method to process the images of grass. As the Image segmentation is the basis for image recognition, the choice of the methods of the color model and the threshold are the keys that affect the quality of image. As the result, before this study set up a reliable and objective Image processing and analysis method, we have to use the appropriate the Color model and the Threshold method first, to get the best results of image processing and analysis.

To filter and assess objectively the best method of Image segmentation from the Color models and the Threshold method that provide by the image analysis software ImageJ, this study use the method of filtering below to assess:

(1)With the different conditions of indoor-shooting, statistics and analysis the standard deviation of the greyscale value of the bare soil image; according to the score of the standard deviation, filtering the reliability to recognize the feature of bare soil with the component of the Color models.

(2)With the same conditions of indoor-shooting, statistics and analysis the greyscale value of grass image and the correlation coefficient of the growth days of grass; base on the score of the correlation coefficient, filtering how the color model component to the greenness variation of recognize grass growth images and the sensitivity of recognize grass green feature.

(3)End up with comparing the grass survival rate came out from the image analysis software ImageJ and the value came out from Point frame method, and base on the absolute error, the average absolute error and image segmentation quality to filter the feasibility and the accuracy of image processing and analysis system and image segmentation.

After objectively statistics, filter and assess, we can tell that in the color model Lab, color model component ‘a’ has the lowest standard deviation and the highest correlation coefficient and that shows the color model component ‘a’ has some kind of reliability and sensitivity; with the combination of the color model component ‘a’ with the threshold method to estimates the outdoor grass survival rate, the absolute error value is between 0.057% to 6.052%, and the average absolute error (MAPE) is under 5%, shows that the image processing method established here has a confidence accuracy and also has a practical value to the practices.

並列關鍵字

ImageJ ； Grass survival rate ； Color model ； Threshold method

參考文獻

4. 江介倫，黃國禎，邱宏彬，黃瀞瑩，2009，「台灣常用覆蓋草類覆蓋率之研究」，農業工程學報，第55卷，第4期，第90-99頁。

12. 林信輝，2009，「植生工程相關名詞釋義」，水保技術，第4卷，第2期，第138-142頁。

28. A.S. Laliberte., A. Rango, J.E. Herric.,and Ed L. Fredrickson, L. Burkett.,2007,“An object-based image analysis approach for determining fractional cover of senescent and green vegetation with digital plot photography ,”Journal of Arid Environments, Vol. 69, pp.1-14.

31. Ery Arias-Castro., David L. Donoho.,2009,“ Does Median Filtering Truly Preserve Edges Better Than Linear Filtering?,” The Annals of Statistics, Vol. 37, No. 3, pp. 1172-1206.

32. ImageJ User Guide, Available at: http://rsbweb.nih.gov/.