Casting positioning method based on vision

The invention provides a casting positioning method based on vision. The casting positioning method comprises the following steps of casting image acquisition, target detection SSD network model processing, key point searching, circle fitting and hand-eye calibration, and casting positioning completing. According to the method, the target detection SSD network model is used, the casting target inthe image can be rapidly and accurately recognized, the circle center position of the casting can be rapidly calculated in combination with a traditional image processing algorithm, positioning of thecasting is completed, and the casting positioning speed and precision are improved. 本发明提出一种基于视觉的铸件定位方法，包含以下步骤：采集铸件图像、目标检测SSD网络模型处理、关键点查找、进行圆拟合、手眼标定、完成铸件定位。本发明使用目标检测SSD网络模型，可以快速、准确的识别图像中的铸件目标，并结合传统的图像处理算法，可以快速的计算铸件的圆心位置，完成铸件的定位，提高了铸件定位的速度和精确度。