基于GPU的Prewitt算法实现及其在探地雷达中的应用

GPU-Based Parallel Prewitt Algorithm Implementation and Its Application on GPR

摘要: 图像的边缘是图像的重要特征之一，边缘检测是提取图像特征的重要手段。GPU-CUDA并行技术作为当前最热门的高性能处理技术，是并行Prewitt边缘检测算法实现的首选。由于常规的基于CPU的Prewitt算法计算量大、耗时，其应用受到很大限制。为了提高算法的效率，文中应用GPU-CUDA技术实现Prewitt算法及快速成像，获得了较高的加速比。最后，通过将实测探地雷达数据转换成灰度图像数据，并实施基于GPU的Prewitt算法处理的方法思路，对实测探地雷达数据进行处理。试验结果表明该算法不仅运行效率高，而且在突出有效异常，提高目标体的识别能力方面取得实效。

Abstract: The edge of image is one of the important features of the image. Edge detection is an important means to extract image features. GPU-CUDA parallel technology, as the most popular high-performance processing technology, is the best choice for parallel Prewitt algorithm implementation. Since conventional Prewitt algorithm based on CPU is computationally intensive and time-consuming, its application is very restricted. In order to improve the efficiency of Prewitt algorithm, GPU-based parallel Prewitt algorithm and fast imaging algorithm are applied to get higher speedup. Finally, the method is proposed by turning the GPR field data into gray-scale image data, then implementation of GPR field data processing with the Prewitt algorithm based on GPU. It is proved that the algorithm is not only of high efficiency, but also effective to improve target identification capability.