作　　者： (郭珂）; (仝明磊）;

机构地区： 上海电力学院电子与信息工程学院,上海200090

出　　处： 《仪表技术》 2017年第8期18-20,23,共4页

摘　　要： 以常用的行为识别数据集作为研究对象,利用条件神经域模型进行单人行为识别。首先用鲁棒自适应视觉背景提取算法提取特征,然后使用质心对齐方式截取目标区域并转成一维向量,最后将特征向量进行实验训练与测试,并将测验结果与隐动态条件神经域和支持向量机算法识别结果相对比。结果显示,条件神经域模型算法在识别率和稳定性方面都优于另外两种算法。 We take the common data set for recognition behavior as the experimental object and put forward a kind of method that uses the neural field model for single behavior identification. First of all,the feature is extracted by the robust adaptive visual background extraction algorithm,and then the centroid is used to intercept the target region and then converted into a one-dimensional vector. Finally,the feature vector is used for experiment training and testing. The results of the test are compared with the results of the hidden dynamic conditions and support vector machines. The conditional neural field model is superior to the other two algorithms in terms of recognition rate and stability.

关 键 词： 行为识别 条件神经域模型 隐动态条件神经域模型 支持向量机模型