To solve the problem that the existing glass bottle detection algorithm only aimed at the smooth surface with no pattern, a glass bottle texture area detection algorithm based on the POBT (phase only based transition) is proposed in the paper. The method first calculates the glass bottle center axis according to the bottle body gray projection, then obtains defect image by removing all the patterns through POBT, which including the process of nomalization in frequency domain to eliminate low frequency component of the glass texture area. At last, it extracts defects by using adaptive probability refined thresholding segmentation on the POBT image and compares the segmentation method with three conventional methods. The experiment result shows that aimed to the glass bottle glass texture area, the method can achieve highprecision detection with rapid speed of the glass bottle texture area.