【摘要】 根据三色测温法原理,利用彩色CCD摄像机所摄取小钢坯图像中的彩色分量,在数字图像处理技术的基础上运用GHM多小波网络(MWNN)对实验数据进行处理,与BP网络处理方法进行比较.实验证明基于GHM多小波神经网络(MWNN)的多光谱辐射测温法在对铸坯温度的测量(温度测量范围为900~1200K)较BP网络训练次数减少一半,训练时间缩到了BP网络的四分之一,并提高了网络收敛速度.更多还原

【Abstract】 In this paper,by applying the digital image processing technology based on the three-color method and GHM wavelet neural networks and applying the color billet images photographed by high temperature CCD the experimental data are processed.Then BP networks are compared with GHM networks which depend on the billet temperature(from 900 to 1200k).The result shows that the multi-wavelet neural network’s precision is better than that of BP networks.更多还原