应用BP神经网络重建物体三维面形

Three-dimensional shape reconstruction based on BP neural network

基于结构光三角投影原理,将神经网络模型引入到物体的三维形貌测量中。BP人工神经网络用于处理变形条纹图像的强度信息,以获得对象的三维形貌信息。该方法通过获取不同高度的数据样本集来训练设计好的BP神经网络模型,直接建立条纹图案的强度分布与对象高度之间的映射关系,并完成对象的三维测量,即便在投影系统参数不确定或变形条纹质量较差状况下,也可以较好的重建。将模型训练好之后,该方法将变形条纹按照神经网络输入形式导入模型就可以直接映射到高度,大大简化了结构光三维测量的计算过程,缩短了测量时间。计算机仿真及实际实验都能重建物体的三维形貌,验证了神经网络方法的可行性。

Based on the principle of structured light triangle projection,the neural network model is introduced into the three-dimensional shape measurement of the object. The BP artificial neural network is used to process the intensity information of the deformed fringe image to obtain the three-dimensional topographical information of the object. The method trains the designed BP neural network model by acquiring data sample sets of different heights,directly establishes the mapping relationship between the intensity distribution of the stripe pattern and the height of the object,and completes the three-dimensional measurement of the object even if the parameters of the projection system are uncertain. Or the deformation of the deformed stripe is inferior,it can also be better reconstructed. After the model is trained,the method can directly map the deformation stripe to the height according to the input form of the neural network,which greatly simplifies the calculation process of the three-dimensional measurement of the structured light and shortens the measurement time. Both computer simulation and actual experiments can reconstruct the three-dimensional shape of the object and verify the feasibility of the neural network method.