自然场景下成熟水果的计算机视觉识别

研究在自然场景下成熟水果的识别技术。比较HSI颜色空间转换的几种方法：球体法、柱体法和双锥体法等。球形：HSI颜色转换系统在本研究中产生较少的奇异数据，被采用。将图像的RGB值转换成色调、饱和度和亮度值，并将色调和饱和度结合产生融合图像，以消除地表、天空等背景的影响。利用Otsu算法自动获取分割阈值，提取目标区域。将提取的区域形态特征进一步分割图像，去除与颜色类似的枯叶和其他背景，提取成熟水果区域，分割效果显著。

Recognition of Mature Fruit in Natural Scene Using Computer Vision

A recognition technique of mature fruit in natural scenes was researched in this paper. First, several color models of HSI system were compared including sphere, cylinder and double cone model etc. The sphere model was selected because of lesser erratic data than others in this research work. The color image was translated from RGB system to HSI system, and the hue, lightness and saturation images were acquired. In order to eliminate the influence of background of soil and sky, the hue and saturation data was combined and produced fusion image. The fusion image was segmented by Otsu operation. As the blasted leaves and other background had similar color as mature fruit, several shape features were extracted to segment object from background. The segmentation result showed this method had a very good effect.