1998 Volume 60 Issue 6 Pages 79-87
This paper intends to develop algorithms capable of discriminating yellow-green apples (var. Orin) in color images, and recognizing individual apples by separating apples connected each other in binary images.
To discriminate yellow-green apples, a neural network model of color image processing was used to distinguish apples, leaves, branches and background. The experiments carried out on yellow-green apple images under the various lighting illustrate that in front of light over 80% (as a target) of the pixels of almost every apples was correctly segmented, but in back of light it was difficult to discriminate apples from leaves when brightness and saturation on apple surface was less than 90 and 30, respectively. The sky and most of branches were correctly segmented.
To separate apples connected each other in binary images, two methods were proposed and tested. The first was multi-threshold processing of brightness on apple surface. The second combined the distance transformation and expansion techniques which was proven to be applicable from the results of which 89% of apples connected were correctly separated and recognized.
