基于步态触觉信息的图书馆智能机器人异常状态检测系统

Library intelligent robot abnormal state detection system based on gait tactile information

针对当前图书馆智能机器人步态识别准确率低,导致异常状态检测效果差的问题,提出基于步态触觉信息的图书馆智能机器人异常状态检测和分类。采用基于局部空间信息加权的K-means算法对静态步态图像进行分割处理,分别构建基于改进K-means的CNN网络模型和基于时域注意力的3D残差网络模型,通过这两个模型对静态、动态步态进行特征提取和识别。实验结果表明,对比于SVM分类器,改进K-means算法的CNN网络模型静态步态识别准确率高达98.7%;3D-CNN模型的动态步态分类准确率为99.72%,均高于其他分类模型。最后结合两种算法进行异常状态检测发现,本算法的分类准确率、敏感度和特异性分别为95.42%、95.53%、94.37%。综合分析可知,提出的算法能够实现静态动态的准确识别和异常状态检测,具有一定有效性。

In view of the problem of poor abnormal state detection effect,the abnormal state detection and classification based on gait tactile information is proposed.The K-means algorithm based on local spatial information was used to segment the static gait images,and the CNN network model based on improved K-means and the 3D residual network model based on time-domain attention were constructed to extract and identify the static and dynamic gait.The experimental results show that compared with SVM classifier,the improved K-means algorithm achieves 98.7%,and the dynamic gait classification accuracy of 3D-CNN model is 99.72%,which is higher than other classification models.Finally,by combining with the two algorithms,the classification accuracy,sensitivity and specificity were 95.42%,95.53%and 94.37%,respectively.Comprehensive analysis shows that the proposed algorithm can realize the accurate identification of static dynamics and anomaly state detection.