Target Identification and Growth State Distinguish of Tomato

The surface color of the mature tomato has little difference with the background color or the entire fruit surface color has inconsistent under the nature growth condition. This fruit image segmentation is very difficulty with the color threshold or the segmentation result is the serious error of the goal outline. According to the near-infrared spectrum and the obvious spectrum respectively effective biology information, the revision and the complement of the harvesting goal outline can been obtained by using the multi-sensor image fusion technology. Taken harvesting tomato into account as an example, based on mutual information best threshold value iteration automatic optimization segmentation (MI-OPT) and the morphology restructuring watershed segmentation are applied to the multi-sensor fusion image of the same target. The shape feature (circle shape, concavity, area ratio) and texture feature (uniformity) are extracted respectively on the basis of morphological process of binary image. And flow of tomato recognition and its growth states´ distinguish are determined according to selected feature quantities. Keywords: Image segmentation; multi-spectral images; image fusion; mutual information; target identification.