Grape Detection By Image Processing

We propose in this paper a new method for grape detection in outdoor images. Zernike moments are used for the description of the grapes shapes. Color information is also exploited. A support vector machine is used for the learning and recognition steps. This method is tested on real images with different grabbing conditions by using a manually made ground truth. The proposed method permits the recognition of grapes in 99% cases with very few samples used in the learning step