基于变分水平集的图像分割防堵仓技术研究

针对煤矿井下噪声高、分辨率低、照度不均的特殊环境，建立煤块图像分割模型，实现对大煤块的准确检测，预防堵仓事故发生。新模型基于变分水平集算法，对GAC(Geodesic Active Contour)和C-V(Chan and Vese)模型改进，融合轮廓和区域模型对煤块视频图像信息进行分割，通过求偏微分方程的稳态解，解决能量模型的最优解求取问题，利用半点离散化差分方法完成数值计算，有效提高计算精度、拓扑结构自适应性、抗噪能力，降低了光照敏感性。实验结果表明，新模型在煤矿井下复杂环境有良好的鲁棒性，实时性高，在减轻了矿工人力筛选大煤块负担的同时极大提高筛选准确率。

Research on prevention of blocking bunker by image segmentation based on variational level set

Due to the special environment of high noise,low resolution and uneven illumination in coal mine,a coal image segmentation model was established to achieve an accurate detection of large coal,and prevent the occurrence of blocking bunker accident.The new model based on variational level set algorithm improved the GAC and C-V model to segment the coal image.The algorithm fused contour and region model,and the optimal solution of the energy model was solved by solving the steady state solution of partial differential equation and the numerical calculation used the method of the discrete slightest difference.The new method effectively improved the accuracy of calculation,topology adaptive capacity,anti-noise ability and reduced the light sensitivity.Experiment showed that the new model had good robustness in the complex environment of underground coal mine and higher real-time performance,reduced the burden of large coal screening and greatly improved the screening accuracy.