边缘检测是提取图像特征的重要手段,已在许多领域得到广泛应用。传统拉普拉斯边缘检测算子不具有Sobel和Prewitt等边缘检测算子的图像平滑功能,对噪声的响应敏感,会误将噪声作为边缘。文中通过求解二阶梯度的拉普拉斯变换,得到了拉普拉斯边缘检测算子,探讨了将传统拉普拉斯边缘检测与二维高斯函数结合,通过一维卷积完成改进拉普拉斯边缘检测的优化算法;同时,在已建立的QT-CUDA并行平台上开发完成了改进拉普拉斯边缘检测模块,并集成到探地雷达精细解释软件系统。对实测探地雷达数据进行处理的结果表明:该算法不仅运行效率高,而且在突出有效异常、提高探地雷达目标体的识别能力方面取得实效。

Edge detection is an important mean to extract image features, and it has been widely used in many fields. Unlike Sobel and Prewitt operator, the traditional Laplacian edge detection operator does not have the function of smoothing image, and it is sensitive to the noise. Thus the noise may be mistakenly considered as edge. By solving the Laplacian transform of second spatial gradient, an improved Laplacian edge detection operator is obtained. The traditional Laplacian edge detection is combined with two-dimension Gaussian function, and the improved Laplacian edge detection optimization algorithm is carried out by the one-dimension convolution. At the same time, the improved Laplace edge detection module is developed based on the QT-CUDA parallel platform, and integrated into the ground penetrating radar (GPR) fine interpretation software system. The experimental results by processing the GPR field data show that the algorithm is not only efficient, but also effective in highlighting effective anomalies, thus improving the recognition ability of GPR targets.

TP391.41

广东省与教育部产学研结合资助项目