目的 为了满足药品缺陷检测系统的具体要求， 需要有更好的图像分割效果。 方法 在二维Otsu法的基础上， 提出 新的阈值分割输出函数， 并改进传统算法， 使运算量大幅度降低。 结果 理论分析和仿真实验结果表明， 该方法更能顾及药品图像边缘细节， 具有好的分割效果， 且分割速度也能满足实时检测要求。 结论 该改进算法用于铝塑泡罩药品包装缺陷检测更具有适用性和实时性。

Objective In order to meet the specific requirements of drug defect detection system, there is a need to have a better image segmentation result. Methods On the basis of two-dimensional Otsu method, a new threshold segmentation output function was proposed by improving the traditional algorithm, which significantly lowered the computational complexity. Results Theoretical analysis and simulation results showed that this new method could cover the drug image edge details, had improved segmentation accuracy, and the speed of segmentation met the requirement of real-time detection. Conclusion The improved algorithm has applicability and real-time performance in defect detection of aluminium-plastic drug packaging.