Tornadoes represent an extremely destructive form of convective weather, posing significant threats to human activities and the safety of life and property. Consequently, the identification and early warning of tornado-producing thunderstorms are essential. Hook echoes frequently appear in the radar reflectivity imagery of such storms. To improve the accuracy of hook echo identification, an algorithm based on radar reflectivity data was proposed. Firstly, the radar reflectivity images were binarized using a 45 dBZ threshold to isolate the intense echo regions, upon which edge detection is subsequently performed. Secondly, Bézier curves were employed to fit the echo edges and identify curvature inflection points. To further confirm the hook echo signature and exclude interference from other echo patterns, the reflectivity gradient feature and the hook echo width feature were extracted. Finally, interference from V-notches was eliminated. The algorithm was validated and analyzed using single-site radar data from Nantong, Jiangsu Province, China, on June 10, 2023, and NEXRAD data from station KUDX in the United States on June 29, 2018. The results indicate that the algorithm achieves a critical success index (CSI) of 0.7 and can reliably identify hook echoes. This method is considered to hold reference value for the early warning of supercell tornado disasters.