Abstract: A novel feature extraction and image processing method for narrow gap weld seam is proposed,which can effectively overcome all type of disturbance and reliably extract the seam feature points.First,a microscopic vision system with the line structure light is designed that can detect the seam image.Then,a method based on pixel point statistics and adaptive morphological contraction is proposed to segment the structured light stripe.After that a straight line and curve fit scheme are used to accurately gain the centre line of stripe.Finally,based on the distance search the characristic points of the seam is obtained by using the algorithm proposed in this research.The effectiveness of these methods is verified by some interesting results.