According to the requirements of non-destructive testing of the anchor bolts in the anchoring engineering, an intelligent detection method of anchor bolts mainly based on correlation function is proposed in this paper. This method uses the nearest neighbor clustering algorithm to automatically pick out the most suitable signals from reflected signals which were collected from the target bolt instead of the traditional manual selection method, and then inputs the selected signal into a proposed hybrid model composed of generalized cross-correlation and Hilbert transform algorithm, so as to automatically locates the mutation points of the signals and then calculates the length of the anchor bolt and the locations of the grouting defects. The experimental results show that the proposed detection methods are more accurate and convenient in practice than the traditional detection methods.