【摘要】 提出利用物体的相位特征联合神经网络的方法对透明半透明三维物体进行识别.首先利用波长扫描数字全息技术和数字再现技术提取物体的相位特征,然后将物体的这些相位特征作为学习模式训练一个BP神经网络,最后利用训练好的网络对三维物体进行识别.实验表明,对于具有小尺度变化的透明半透明三维物体识别,该方法的正确识别率为100%.更多还原

【Abstract】 A new approach based on phase features combined with neural network model is proposed for recognizing 3-D objects. The phase features of an object were extracted by wavelength-scanning digital holography and numerical reconstruction technique. A BP neural network with one hidden-layer trained by reconstructed images of three pyramids was used to recognize other pyramids with some variance, and the correct recognition rate of these pyramids is up to 100%. The simulation results demonstrate that the method is effective.更多还原