基于BP神经网络的注射器针头缺陷检测方法

为了实现倒装和弯钩次品针头的自动检测,提出了一种基于BP神经网络的注射器针头合格性检测方法。该方法首先对针头图像进行去噪、目标分割和针头轮廓提取等预处理,其次采用边界区域不变矩法和针头边缘曲率法提取针头特征,然后用合格针头、弯钩针头和倒装针头样本的特征对设计好的BP神经网络进行训练,最后利用训练好的BP神经网络实现注射器针头的合格性检测。通过大量真实针头的合格性检测实验,验证了本研究所提出方法的有效性,可用于实际生产中。

A Syringe Needle Eligibility Detection Method Based on BP Neural Network

In order to achieve automatic detection of flip and hook defective needles, a syringe needle eligibility detection method based on BP neural network was proposed. Firstly, there are several preprocessing steps including the needle image de-nosing, needle target segmentation and needles contour extraction. Next, needles feature extraction followed by boundary region invariant moment method and needle edge curvature. Then, the designed BP neural network is trained by the samples of the qualified needles, bent needles, and inverted needles. Finally, the quality of the needle is tested by the trained BP neural network. Through a lot of real needle detection experiment, the proposed method is effective and can be used in actual production.