【摘要】 为了有效地解决在线高速实时检测这个关键性问题,本研究主要从软、硬件两个方面进行,将制约系统实时性的缺陷图像的识别过程分离出来,交由DSP芯片进行专门的处理,同时采用了兼顾识别效率及识别准确性的支持向量机的二级分类器对带钢的缺陷图像进行识别.在该硬件检测系统下对缺陷图像的正确识别率达到98%,缺陷图像的识别时间可控制在10 ms以下.通过理论上的分析和试验的测试证明所搭建的先进的TMS320C6416 DSP图像处理平台能够很好地满足实际生产线上的带钢表面缺陷的实时检测系统在处理速度和精度上的要求.更多还原

【Abstract】 Taking account of both software and hardware,the effective real-time high-speed inspection as an on-line problem to solve was investigated.The recognition process of strip surface defect images restricting the real-time aspect was separated,then it was processed specially by a DSP chip,and simultaneously recognized by a secondary classifier of support vector machine that combines the recognition rate with accuracy.By such an inspection system,the correct recognition ratio of defect images comes up to 98% with time controlled lower than 10ms.It was proved theoretically and experimentally that the advanced TMS320C6416 DSP image processing platform thus developed can meet well the actual requirement of on-line inspection for strip surface defects in both speed and accuracy.更多还原