The current technology for defect detection of composite curved components has problems of unintuitive detection results and low detection efficiency. To address these issues, a three-dimensional defect imaging method based on the phased array ultrasound is proposed. Three-dimensional scanning technology is used to obtain the surface model of the curved surface, and the parallel section method is used to design the detection trajectory. Then, the sequence of two-dimensional ultrasound data is collected by the phased array ultrasonic wheel probe. The polynomial method is used to fit the detection trajectory curve function. The normal vector and space coordinates of the two-dimensional ultrasound data are calculated according to the scan step length and the sequence relationship. The ultrasound point cloud set is generated. Finally, the voxelization down-sampling method is used to realize the three-dimensional imaging of the internal defects of the composite material. Experimental results show that the average error of defect imaging results between ultrasonic detection and CT detection is 1. 14 mm, which proves that the proposed method can quickly reconstruct the location, shape and size information of the defects, and accurately characterize of the internal defects of the composite material curved sample.