Abstract: The safe operation of gas-insulated metal-enclosed switchgear (GIS) is seriously threatened by density non-uniformity and concentrated internal defects of epoxy composite insulators. There are deficiencies in the ultrasonic detection methods, such as non-intuitive defect detection and low efficiency. Therefore, this study develops an ultrasonic automatic imaging system for internal defects of epoxy composite insulator materials based on the principle of ultrasonic reflection. Firstly, it uses wavelet analysis to denoise non-repetitive ultrasound echo signals. Then, feature fusion recognition is performed on the time differences between adjacent peaks of the echo waveform, voltage peak, and zero signal time between adjacent voltage peaks to obtain the sample defect location-ultrasonic feature dataset. Finally, the automatic ultrasonic imaging of epoxy composite insulation defects is realized. The study uses the developed system and the traditional detection methods respectively to detect and compare the epoxy composite insulation samples with density non-uniformity and cracks of different depths and orientations. The results show that the local density error between the system and the mass volume method is less than 4.8%. The detection results of internal concentrated defects are consistent with the original appearance of defects, and the detection efficiency of the system is nearly 6 times higher than that of the traditional artificial ultrasonic detection. This detection system combines the detection of concentrated internal defects and density non-uniformity in insulation materials. Compared with the traditional manual detection methods, it offers significant advantages in terms of detection efficiency and intuitiveness in insulator delivery testing and defect localization after faults.