基于OpenMP并行技术的多核环境下轮对图像处理方法研究

Wheel set image processing method under the multi-processor environment based on OpenMP parallel technology

基于光电图像的轮对磨耗在线检测过程中,轮对图像处理的速度对在线检测的实时性有重要影响。针对轮对图像的处理速度不能够满足轮对高速在线检测要求的问题,利用多核处理器的计算优势,对基于Open MP并行技术的轮对图像处理方法进行了研究。通过分析轮对图像的特点,对不同类型的轮对图像采用了不同的分割方法,减小了轮对图像分割环节对后期处理的影响;提出了采用多图并行处理的编程模式,设计了并行优化方案,利用Open MP并行技术对轮对图像的处理进行了并行优化;并将Open MP四线程与单线程对轮对图像的处理结果进行了对比分析,验证了轮对图像处理加速效果。研究结果表明,多核环境下,轮对图像处理通过Open MP并行优化,获得了高于2倍的加速比,可满足车辆轮对检测速度达到80 km/h的在线检测要求。

During the online detection of wheel set wear based on optoelectronic image, the speed of wheel set image processing has significant impact to the real-time of online detection. Aiming at the problem that the wheel set image processing speed can not meet the requirement of high speed wheel set online detection, the calculation advantage of multi-core processors was used and the wheel set image processing method based on OpenMP parallel technology was studied. Through the analysis of the characteristics of wheel set image, according to the different type of wheel set image, different segmentation method was used and the influence of wheel set image segmentation step to the post- processing was reduced; the programming mode of multi image parallel processing was proposed and the parallel optimization scheme was designed, the OpenMP was used to realize the parallel optimization of the processing of wheel set image. Then the image processing effect of OpenMP four threads and single thread was compared to verify the processing acceleration effect of wheel set image through the experiment. The research results indicated that, based on multi-core, the parallel optimization of wheel set image processing by OpenMP has improved the speed more than 2 times and met the requirement of wheel set online detection under 80 km/h speed.