双目视觉机器人遮挡目标识别研究

Research on Recognition of Occluded Target for Binocular Vision Robot

摘要:

随着《中国制造2025》的提出,中国的装备生产和应用的信息化与智能化提上了日程。而双目视觉机器人恰是实现这一目标的执行者。双目视觉机器人识别过程，受实际工况和自身目标识别算法的影响，具有盲目性、非穿透和波动等特性， 针对该复杂过程，提出了UWB遮挡目标识别算法。该算法参考无线电波的运动学方程，结合UWB(Ultra Wideband)无载波通信技术，构成遮挡目标识别系统。并通过UWB标定原理对遮挡目标识别系统进行标定实现对遮挡目标的精确识别，通过实物运行，误差控制在7.1%，满足位置偏差小于26.3%的设计要求，验证了该方案的可行性和有效性。该研究对双目视觉机器人的双目视觉目标识别算法结合其他目标识别算法组成复杂的目标识别算法网络是一次有益的尝试。

Abstract:

With the proposal of Made in China 2025, the informationization and intellectualization of equipment production and application in China have been put on the agenda. Binocular vision robot is the executor of this goal. The recognition process of binocular vision robot is influenced by actual working conditions and its own target recognition algorithm. It has the characteristics of blindness, non-penetration and fluctuation. Aiming at this complex process, a UWB occlusion target recognition algorithm is proposed. The algorithm refers to the kinematics equation of radio wave and combines Ultra Wideband (Ultra Wideband) carrier-free communication technology to construct an occluded target recognition system. The UWB calibration principle is used to calibrate the occluded target recognition system to achieve accurate recognition of the occluded target. Through physical operation, the error is controlled at 7.1%, which meets the design requirements of position deviation less than 26.3%, and verifies the feasibility and effectiveness of the scheme. This study is a useful attempt to combine the binocular vision target recognition algorithm of binocular vision robot with other target recognition algorithms to form a complex target recognition algorithm network.