Abstract: With its large field of view, GWAC can record hundreds of meteors every day. These meteors are valuable treasures for some meteor research groups. It is therefore very important to accurately find all of these meteors. To address the challenge of precisely distinguishing meteors from other elongated objects in a GWAC-like sky survey system, we design and implement a meteor candidate recognition algorithm, including the recognizing and morphology analysis of the light curves of the meteor candidates. After processing the images of Mini-GWAC taken in two months, we detect 109000 elongated objects in which more than 90 percent of objects are not meteor. By analyzing α parameter of gaussian fitting curve of meteor target in midnight, we find that the wave crest of most single-peak meteor target changes slowly with the position of image pixel. Among the elongated objects, about 4.1% objects are identified as meteors with high confidence, after applying filters based upon an existence in a single frame, a single peak in the light curves, and a slow variation of the light curves. After manual examination, 85%-87.3% of the 4.1% candidates are consistent with the shape and brightness of the meteor.