A segmentation method was developed for combining an improved adaptive genetic algorithm with a two-dimensional maximum entropy algorithm based on the image features of ultrasound infrared thermography detection for accurate and rapid segmentation of the target defect region to recognize defects of defect recognition in images. First, the infrared image was processed to obtain a denoised image. Next, the image was divided into the target and background regions using a two-dimensional maximum entropy algorithm; the segmentation speed was improved by combining it with the improved adaptive genetic algorithm. The experimental results showed that this method can effectively filter image noise. Compared with an exhaustive method and the two-dimensional maximum entropy segmentation based on a simple genetic algorithm, the proposed algorithm has better segmentation speed and accuracy.