Canopy Image Segmentation Based On Improved Qtsu Method And Leaf Area Index Estimation Of Ginkgo Biloba

The traditional estimation method of leaf area index (LAI) is complicate and destructive.With the development of information technology, traditional forestry research methods havebeen difficult to meet the needs of the development. Computer image segmentation techniquesare increasingly important in forestry. How to extract and analyze the tree information from theimage data is very important for deeply analysis of trees and the forest in modern forestryresearch. In fact, it is one of the crucial issues in the subsequent image research that accuratelysegment the target plant from complicated forest image. Among all the image segmentationmethods, Otsu method becomes one of the most widely used threshold selection methods ofimage segmentation because of its simple calculation and great adaptability. However, due tothe naturality of the tree images, it is not able to obtain good segmentation results when wesegmented the forestry canopy images using Otsu method.Considering its poor segmentation results of the canopy images, the objective function ofautomatic threshold selection is improved and an improved threshold selection algorithm basedon traditional Otsu method is presented in this paper. It considered both between-class varianceand within-class variance that impact on the image segmentation result. Taking the Ginkgobiloba canopy images of different canopy density as examples, we compared the segmentationresults obtained by traditional Otsu method and those obtained by the improve Otsu method,and the results show:(1) For the images of lower canopy density, the results obtained bytraditional Otsu method and improved Otsu method are similar.(2) For the images of highercanopy density, better segmentation results can be obtained through the improved methodrather than traditional Otsu method.(3) According to the results of lots of canopy imagessegmentation test, it is feasible to segment the canopy images using the improved Otsusegmentation method and better segmentation results can be obtained by the improved methodrather than traditional Otsu method. Using the “Forestry seedling image analysis software”, two data sets of images, whichwere captured by fisheye lens and wide-angle lens, were segmented by improved Otsu methodproposed in this paper. We fitted a model between image information obtain by improved Otsusegmentation method and the leaf area index. The independent variable was the ratio of thesegmented foreground pixels to the total pixels of the image and the dependent variable wasthe LAI. The model evaluation and model checking results show that:(1) The fitting modelwell figured out the relationship between segmented image information and Ginkgo biloba leafarea index.(2) The fitting results of the two image data sets which captured by fisheye lens andwide-angle lens were similar. The wide-angle lens can be used in actual research, which cansave the cast.(3) The presented non-destructive, fast and reliable method of leaf area indexestimation is feasible.