A method for feature points extraction of planar weld seams based on genetic algorithm is proposed. In order to reduce the image noises, we use median filtering method and threshold segmentation method to preprocess welding images. The seed filling method is used for the image segmentation, and the mathematical model of laser stripe skeleton extraction is obtained according to the characteristics of the image. The skeleton extraction method of laser stripe based on genetic algorithm is mainly studied, and the coordinate of center point is extracted with linear scanning method. The Pauta criterion is used during the linear fitting of the skeleton to iteratively eliminate the noise data, and the accurate position of the skeleton and feature points are obtained. The experimental results show that the method can effectively eliminate many noises and the interference of laser stripe width in weld image and can extract the weld feature points quickly and accurately.