针对传统的自动焊线杌采用模版法进行LED微芯片检测速度慢、效果不理想的问题,提出了一种无需模版的检测方法.该方法首先进行图像分割和改进的中值滤波处理,然后采用自适应阈值算法将图像二值化,再利用邻域检测将图像中的微芯片适当加权,最后通过质心定位微芯片.部分可并行化的操作(改进的中值滤波、加权操作)利用GPU并行运行,提高定位速度.实验结果表明,该检测方法鲁棒性好,抗干扰性强,对形状不一、边界模糊和部分残缺的各种LED微芯片都可操作.与传统方法相比,适用范围广,速度和准确度显著改善,提高了自动焊线机的生产效率.%As the traditional method of using templates to LED microchip detection with low efficiency,this paper proposes a detection method without template.First image is divided and improved median filtering processing,secondly to adopt new adaptive threshold algorithm binary image.Thirdly.the image is weighted by the neighborhood detection,finally through calculating quality center localization microchip.Then use GPU makes the whole operation parallelization to accelerate positioning speed.Experimental results show good robustness of the detection method,immunity to interference.Different shapes,boundary ambiguity and incomplete microchip can be applied.Compared with traditional methods,this method was widely applicable,speed and accuracy improved markedly, improves automatic goldwire welder production efficiency.
展开▼
