Non-destructive determination of leaf area in tomato plants using image processing

Publisher Website
Google Scholar
Abstract
The objectives of this study were to validate the use of image processing for the measurement of leaf areas in intact tomato plants and to examine the potential of taut- string bounding area as a parameter for wilting. This was achieved by taking images of 18 tomato plants, viewed from the top, from the side and from an oblique angle. In the case of side and oblique views the images were taken as the plant was rotated through 360° with 45° steps. An average value of the leaf area was then calculated for each view. Regression of true leaf area on image-calculated areas for individual views resulted in a second order polynomial which adequately described the relationship. The regression coefficient, R2, adjusted for degrees of freedom, was 0.98. Multiple regression of true leaf area against areas obtained from two views increased R2 to 0.99. When all three view points were taken into consideration, R2 increased further, albeit modestly, to 0.993. Taut-string boundary provides information about the plant’s compactness as it grows. A boundary was obtained by rotating the image co-ordinate system through 16 different angles. After every rotation the extreme points of the image were determined. These points formed the corners of a bounding polygon that approximated to the taut string boundary. The polygon area was calculated from the co-ordinates of the polygon corners. Regression of true leaf area on taut string image area for individual views resulted in a second-order polynomial which adequately described the relationship. Multiple regression improved the regression coefficient. To test the potential usefulness of taut-string area, two plants were water stressed, then irrigated and monitored while they recovered from wilted to turgid. The change in taut string area detected (50–55%) was sufficient to recommend its use as a meaningful parameter for wilting.

This publication has 6 references indexed in Scilit: