This article proposes a recognition method and storage medium based on workpiece shape feature coding, which achieves automatic recognition and classification of workpiece shapes by generating unique feature codes. The method first uses a structured light projection method to establish a 3D model, and then extracts spin graph features for workpiece shape coding. In the coding process, the key points are set as the curvature maximum points with the maximum distribution in the region, and the threshold value for the minimum distance between vertices is set, finally determining the n curvature maximum points on the surface of the 3D model according to the regional distribution. The spin graph is a local surface description operator that is invariant to translation and rotation geometric transformations and can describe the global features of the workpiece shape. Combining the spin graph features with the point cloud model can achieve accurate recognition and classification of the workpiece shape. The method also designs a supervised hash encoding architecture, which maps the input features to a vector space and generates feature identifiers using a hash function. This method can effectively improve the discrimination and recognition rate of the features, realizing fast and accurate recognition of 3D models.