Three-dimensional measurement of complex object based on wavelet neural network
CCI Yi-Xiang, CHEN Wen-Jing, ZHAO Yue, XU Luo-Peng (Department of Opto-Electronics, Sichuan University, Chengdu 610064, China)
The wavelet neural network has been introduced into the reconstruction of the complex three-dimensional (3D) object based on structured light projection. In the method, the wavelet with time-frequency characteristics and zoom features and the neural network with powerful function of approximation is used to get the continuous approximate function and draw phase distribution of the object. As a result, the wavelet neural network method based on structured light projection needs only one deformed fringe pattern to reconstruct the tested object. Compared with the Fourier transform profilometry, the wavelet neural network without filtering process and with high sensitivity can demodulate more useful phase from the fringe pattern with shadow. Therefore, this method performs better than Fourier transform profilometry in the three-dimensional shape measurement of complex objects. The feasibility of this method is validated by computer simulations and experiment.