针对实际应用中轨面伤损检测系统检测速度较慢的问题, 在保证检测精度的前提下, 结合轨面图像特点及软件工程的思想, 提出了面向算法、编程技术和存储介质3个层面的优化方法。算法优化通过重新设计算法流程, 合理取舍步骤, 实现算法到CPU的高效映射; 编程技术优化使用多线程编程, 通过并行运算充分利用处理器的多核优势; 存储介质优化通过使用读写快、质量轻、能耗低、体积小的固态硬盘进行图像读写, 有效地提升了硬件效率。实验结果表明, 优化后平均每幅轨面图像检测耗时由17.94 ms降低到仅8.33 ms, 速度提升了53.57%, 在分辨率为1 mm的精度下换算成车速约为207 km/h, 可以满足铁路轨面伤损在线检测需求。

Under the premise for guaranteeing the accuracy of detection, combining with the characteristics of the rail surface image and the software engineering concepts, 3 optimization methods were presented for the issue of slower speed in rail surface defects detection, including the aspects of algorithm, programming technology and storage medium. According to redesign of the algorithm flow, the efficiency of the algorithm mapping to CPU was improved; the multi-thread programming was used to make full use of the multi-core strengths of the CPU; the efficiency of the hardware was improved by using solid state drive (SSD) to read and write images, the SSD has the characteristics of fast reading and writing, light weight, low energy consumption and small size. The experiment results demonstrate that the average time consuming decreases from 17.94 ms to 8.33 ms per picture after optimization,the speed improves by 53.57%. That means the speed of train is about 207 km/h in 1 mm accuracy resolution and the system can satisfy the requirements of online detection of rail surface defects.