基于神经网络的硬币识别与分拣系统

Research on Coin Denomination Recognition Based on Neural Network

提出基于神经网络的硬币识别与分拣系统的设计方法。硬币传输过程中,根据金属特性的不同利用电涡流技术完成真伪识别;硬币原始图像采集后经灰度变换、中值滤波降噪,由胡氏几何不变矩提取硬币图像特征,将该特征作为三层BP神经网络的输入,实现硬币面值识别。实验表明该设计的硬币识别与分拣系统,能够完成对硬币真伪及币值的分辨,识别速度快,识别精度高,具有应用价值。

The coins recognition and sorting system based on BP-Net is put forward.Using electric eddy current sensors for non-contact detection and putting the non-electric quantity into electrical signals,false coin can be picked out.The original image of the coin was transformed by grayscale and denoised by median filtering,its features were extracted by Hu's geometric invariant moment,used as the input of three-layer backpropagation neural network,and the face value of the coin could be identified.Experiments show that the designed coin recognition and sorting system can recognize the authenticity and face value of the coins.The system has fast recognition speed,high recognition accuracy and application value.