The slight sudden changes in the surface of metal object can be highlighted by the distortion of reflection stripes. Therefore, the reflection stripe technique can be applied in the surface inspection of reflective objects. We propose a machine vision detection method for the surface indentation of cold stamping valves based on reflection stripe image. Along the way, stripe image information of cold stamping valves is extracted, and defect features are recognized automatically. A series of preprocessing methods, such as noise filtration and multi-scale Retinex algorithms, are adopted to improve the image quality. Characteristic parameters, such as fringe-centerlines, sum of pixels, and projection vectors in child windows, are selected to reduce the computation complexity and improve the robustness of the computing system. The experimental results show that this detection method for the surface indentation of cold stamping valves based on reflection stripe image has high accuracy and high efficiency. This method can achieve effective identification of subtle indentation on the surface of cold stamping valves to an accuracy of 0.1 mm, and detection time efficiency (one valve takes 2 seconds) meets the online detection demand for the cold stamping valve production line.