基于神经网络算法的机器人定位技术研究

机器人定位抓取工件时,正确的选择工件特征参数是机器人能否准确获取工件抓取点,进而对工件进行抓取的成败关键。在研究了图像处理技术的基础上,提出了利用神经网络非线性处理能力解决工件特征选择和特征提取过程中存在的非线性问题。在神经元的训练中,通过使用改进的Hebb学习规则克服了传统学习模式下的权值无限制增长而不收敛的问题,提高了特征的识别度和特征提取的准确性,使机器人能够实现对工件的准确抓取。

Research on Localization Technology of Robot Based on Algorithm of Neural Network

When the robot is positioned to grab workpiece,the correct of choosing workpiece feature parameters is the key to whether the robot can obtain the grabbing point of workpiece accurately and grab the workpiece. Based on the method of image processing, neural network nonlinear processing ability was proposed to solve the nonlinear problems in workpiece feature selection and feature extration. In the training of the neurons, by using the improved Hebb learning rule, the problem of unrestricted growth of weights and no convergence in traditional learning mode was overcome.The recognition degree of the feature and the accuracy of feature extraction are improved ,which makes the robot realize accurate scraping of the workpiece.