基于多源机器视觉信息融合的番茄目标匹配

摘要

There is the uncertainty problem of characteristic match in the two dimensional images process of tomato acquired by robot binocular vision in natural growth condition. With the biological information of near-infrared spectrum and visible spectrum, the tomato's effective centric spots of multi-sensors image fusion are obtained in the binocular matching search, and the epipolar constraint and the unique constraint are adopted to do the regional bi-directional matching.. The results indicated that the match method may realize the only match of the target and had the high accuracy rate.%在番茄自然生长条件下利用计算机双目视觉获取的二维图像其处理必然会涉及到特征匹配不确定问题.该文利用近红外光谱和可见光谱各自有效的生物信息,在双目匹配搜索中,提取多源视觉融合图像的番茄有效形心点,采用极线约束和唯一性约束进行区域相关双向匹配.试验结果表明,基于此匹配方法可以实现果实的唯一匹配,准确率较高.