A dual-camera three-dimensional (3D) imaging system is established by using the fringe projection technique to measure objects with large slopes an d/or discontinuous surface.A DLP projector projects the generated sinusoidal fri nge pattern images onto the measured objects surface.The fringe patterns are deformed with regard to object surface and captured by two color cameras from di fferent views.The absolute phase of each pixel can be calculated from the captured fringe patterns by using the optimum three-fringe numbers selection method.The relationship between the absolute phase and depth is estab lished,so two point cloud data from different coordinate systems are obtained.An improved iterative clos est point (ICP) algorithm is proposed,and the obtained point cloud data can be transformed into the same coordinate system by rejecting noises and invisible data automatically at each iterative step.Experimental results on measuring 3D shape of objects show the accuracy and availability of the developed 3D imaging system.The maximum measurement e rror is 0.072mm in measurement range.