论文部分内容阅读
针对远程实时识别挖掘机图像这一问题,研究了基于不变矩及BP神经网络的识别分类方法。首先,获取挖掘机铲斗的原始图像,然后对其进行处理得到二值化图像。以提取出的七个不变矩特征作为样本,输入三层BP神经网络。通过训练,对不同位置的铲斗图像进行识别分类,判断出挖掘机的工作状态。实验表明:该方法识别率较高,提取到的铲斗目标信息及姿态图像信息,对后续的视觉伺服控制研究有一定帮助。
For the problem of remote real-time recognition of excavator images, the recognition classification method based on invariant moment and BP neural network is studied. First, the original image of the excavator bucket is taken and processed to obtain a binarized image. Taking the seven invariant moment features extracted as samples, we input three-layer BP neural network. Through training, the different positions of the bucket image recognition classification, to determine the working status of the excavator. Experiments show that this method has high recognition rate, and the extracted bucket target information and attitude image information are helpful to the subsequent research of visual servoing control.