多媒体环境下船舱监控视觉图像敏感区域标注算法

Sensitive area labeling algorithm for visual image of cabin surveillance in multimedia environment

在船舱监控视觉图像的敏感区域标注中,针对敏感区域聚类标注算法准确率较低的问题,在多媒体环境下,提出了一种船舱监控视觉图像敏感区域标注算法。使用视觉注意模型计算各个区域的敏感度,对敏感区域进行检测。利用K-NN聚类算法对图像样本进行聚类,对出现频率最大的样本类别进行标注。利用图像的SIFT特征对图像的特征点与梯度进行描述,得到敏感区域图像特征后,根据值的大小,排序所有未标注的图像标注词,实现了船舱监控视觉图像敏感区域标注算法。为了检测该算法,利用敏感区域聚类标注算法与该算法进行敏感区域标注准确率的对比实验,证明了船舱监控视觉图像敏感区域标注算法的可行性与准确率。

In order to solve the problem of low accuracy of clustering algorithm in sensitive area labeling of ship cabin surveillance visual image,a new labeling algorithm for sensitive area of ship cabin surveillance visual image is proposed in multimedia environment.Visual attention model is used to calculate the sensitivity of each region and detect the sensitive region.K-NN clustering algorithm is used to cluster image samples and label the most frequent samples.The SIFT feature of the image is used to describe the feature points and gradients of the image.After the image features of the sensitive area are obtained,all the unmarked image annotations are sorted according to the size of the value,and the algorithm of the sensitive area annotation of the visual image of ship cabin monitoring is realized.In order to detect the algorithm,a comparative experiment was carried out between the sensitive region clustering labeling algorithm and the algorithm,which proved the feasibility and accuracy of the sensitive region labeling algorithm for ship cabin surveillance visual image.