Welcome to Chinese Agricultural Science Bulletin,

Chinese Agricultural Science Bulletin ›› 2015, Vol. 31 ›› Issue (35): 236-244.doi: 10.11924/j.issn.1000-6850.casb15080012

Previous Articles     Next Articles

OpenCV-Based Measurement System for Plant Leaf Geometry Parameters Using Android Mobile Phone

  

  • Received:2015-08-03 Revised:2015-11-17 Accepted:2015-08-19 Online:2015-12-18 Published:2015-12-18

Abstract: The study aims to use Android mobile phone to measure the plant leaf geometry parameters, including length, width, perimeter, area, etc., to avoid the weakness of other methods and improve the data collection efficiency. Firstly, the authors obtained the image of leaf and a known size rectangular box. Then, the image correction was conducted to eliminate the distortion. In the correction step, a more robust method of detecting the feature points was used instead of the method of Hough transform to process the rectangular box and got its 4 corner points. According to these points and the corner points of the standard image, the authors mapped the matrix to make the image correction. Finally, the authors calculated the leaf geometry parameters by processing the corrected image. The parameters were calculated based on the leaf contour. There were 3 advantages of using contour: it can avoid the phenomenon of mistaking noise pixels for leaf pixels in the counting pixels method; morphological processing which was used for eliminating the gap and discontinuity of leaf would not be necessary; it was convenient to obtain the geometry parameters because all of them were calculated based on the contour. And there were 2 methods that users could choose to calculate the length and width. In addition, the OpenCV-Android-SDK was used in programing which could process image fast. The results showed that: in the accuracy test, the measurement error of printed regular shape geometry parameters was less than 2%; and in the measurement of real leaves, the error of length and width were less than 2%, the error of perimeter was under 4% and the error of area was below 3%. In the accuracy test, it was faster by using OpenCV in the same image processing algorithm. In test of several phones of different configurations, all phones processed a 2448×3264 resolution leaf image within 3 seconds. Moreover, for the same image, different phones had the same measurement results. The experiment results showed that this system not only had the advantages of easy operation, fast speed, strong universality, but also had a high accuracy on the measurement of leaf geometry parameters.

CLC Number: