Digital image processing based methodology for accurate estimation of leaf length, width and aspect ratio

Image processing for leaf dimension estimation

Abstract

This paper suggests a method for estimating the length and width of plant leaves that is based on image analysis. Accurate measurements of leaf dimensions, including length, width, and aspect ratio, are essential for physiological research, crop development initiatives, and plant phenotyping. A flatbed scanner is used to capture the image, which is then saved in TIFF format and converted to binary. The image is then thresholded. Following thresholding, a bounding box was created, and the leaf's length, breadth, and aspect ratio were approximated using it. In addition to being labour-intensive and time-consuming, traditional manual measures are also prone to human mistake. For calculating leaf linear parameters, image processing techniques offer a quick, automated, and accurate substitute. This study offers a reliable approach for utilising image processing techniques to estimate the length, width, and aspect ratio of leaves. Image acquisition, pre-processing, rotation, segmentation, feature extraction, and dimensional analysis are all included in the suggested method. The coefficient of determination (R2) and root mean square error (RMSE) were analysed to assess the experimental outcomes. Leaf length, width, and aspect ratio have maximum RMSEs of 0.565, 0.275, and 0.195, respectively. In a similar manner, R² values of 0.980, 0.955, and 0.981 are determined for leaf length, width, and aspect ratio, respectively. The suggested methodology's ability to determine the length, width, and aspect ratio of such leaves with a high degree of accuracy is demonstrated by the low RMSE and high R². The methodology's correctness and efficiency are demonstrated by experimental results, which make it a useful instrument for plant sciences and agronomy.