Image processing techniques in the evaluation of droplet size distribution on sprayed leaves

摘要

Four image segmentation techniques are assessed for use in determining the coverage of crop leaves by aerosol droplets after the application of chemicals for plant disease and pest control. The image analysis problem consists of effectively determining the number of droplets as well as the droplet size distribution on the sprayed leaves. Images of sprayed leaves are analyzed using a standard CCD camera mounted on a microscope. The entire image is digitized into 512 scan lines, each consisting of 512 pixels. Each pixel is assigned an integer value between 0 and 128 based on a brightness gray scale. Segmentation of the image is done using different thresholding techniques. The four methods compared are: the entropy method of T. Pun (1981); the moment preserving method of W.H. Tsai (1985); the maximum correlation method of A. Brink and G. De Jager (1987); and the minimum difference method. The usefulness of these methods are evaluated as to their accuracy in representing a bilevel image which can be used to determine the correct droplet areas. The maximum correlation method is recommended as the best of the four methods.