ILBP算子在浓雾天气形势图识别中的应用

Application of ILBP Operator in Recognition of Fog Weather Situation Map

纹理是天气形势图的突出特征,有效地从天气形势图提取并表示其纹理是实现雾型实时在线预报的基础。基于此,提出一种改进的局部二值模式算法,通过调整局部二值模式(LBP)算子中二进制多项式的权重,实现其提取特定方向上纹理特征的目标。将江苏地区2010年—2017年500张浓雾天气形势图作为数据集,采用Chi统计法匹配测试数据与基准数据的相似度进行天气分类。实验结果表明,该算法的准确率、虚警率及临界成功指数分别为0.884、0.15和0.76,均优于LBP算法,具有较高的识别准确性与可靠性。

Texture is the salient feature of a weather situation map.Effectively extracting the textures from the weather situation map and representing them is the basis for real-time online prediction of fog pattern.To this end,this paper proposes an Improved Local Binary Pattern (ILBP) algorithm,which can extract texture features in a specific direction by adjusting the weight coefficients of binary polynomials in Local Binary Mode (LBP) operators.500 fog weather situation maps in Jiangsu Province from 2010 to 2017 are used as the dataset,and the similarities between the test data and the benchmark data are matched by Chi statistical method to classify the weather.Experimental results show that,the Probability of Detection (PoD),False Alarm Rate (FAR) and Critical Success Index (CSI) of the proposed algorithm are 0.884,0.15 and 0.76,respectively,which are better than those of the LBP algorithm.It proves that the proposed model has high accuracy and reliability.