摘要
针对常用的三坐标测量设备在对物体表面采样时造成的局部区域数据残缺问题,提出并试验了一种三维残缺数据的BP神经网络修补方法。该方法首先人裁出一小片数据,然后框选出残缺数据边界附近的样本点集,并加以训练;最后,在给出残缺数据映射于xoy平面的xy坐标情况下,把训练后的BP网络用于残缺区域数据点z值的预测;根据预测值和真实残缺值的对比实验分析表明,此方法具有较高的修补效率和精度,可获得满意的修补效果。
Due to surface reflection property, occlusion and accessibility limitation, certain areas of the object were usually not sampled, leading to holes and undesirable artifacts in the resulting models. To improve the quality of modeling, a new repairing method for 3D incomplete data was proposed and tested. First, cut out an contrived area of the surface points; Second, the sample points were acquired around the incomplete data boundary by using a rectangle frame, and training. Finally, based on the two series data of x y＇s value, which mapped to the plane of XOY, forecasting the value of z＇s value acoording to the BP Network that was trained. Compare the forecasted values and the value actually it is.Experimental results show that the method has food repairing efficiency and precision.
出处
《机械工程师》
2007年第2期32-34,共3页
Mechanical Engineer
