皮带撕裂的视觉检测

Visual Inspection of Belt Tearing

为保证物料输送皮带的运行安全,构建了皮带撕裂视觉监控系统,并基于改进的BP神经网络方法实现了对皮带裂纹的视觉识别。首先,采用Canny算子对皮带裂纹进行边缘提取,并提出了对检测到的边缘进行四周外向扩展的方法,通过增加一个裂纹权重值,使检测到的裂纹边缘向外扩张,从而获得更鲁棒的边缘检测效果;然后,构建改进的BP神经网络预报模型,以皮带裂纹图像的面积及长宽比几何特征量作为网络输入量,采集皮带裂纹图像样本,实现BP网络预报模型的训练及预报结果输出。实验表明:提出的皮带撕裂检测方法是有效的。

In order to ensure the operating safety of conveyor belts, the paper established a belt tearing visual inspection system, in which the belt cracks were realized to be visually identified based on the improved BP neural network method. Firstly, the belt cracks was performed an edge extraction by adopting Canny operator. In addition, a method of extending the detected edge outwards was put forward, in which the crack edges would be extended outward by adding a weight value, so that a more robust edge detection result would be obtained. Furthermore, an improved BP neural network forecast model was established, of which the area and length-width ratio of the belt crack images were regarded as network inputs, in order to take samples of belt crack images and implement the training and forecast result outcomes of the BP network forecast model. The experimental result shows that the proposed belt tearing inspection method is effective.