论文部分内容阅读
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。
The commonly used behavioral recognition dataset as the research object, the use of conditional neural domain model for single behavior recognition. Firstly, the robust adaptive visual background extraction algorithm is used to extract the feature, and then the target area is extracted using the centroid alignment method and converted into a one-dimensional vector. Finally, the feature vector is trained and tested, and the test results are correlated with the hidden dynamic state of the neural domain and support Vector machine recognition algorithm to identify the relative ratio. The results show that the conditional neuromorphic model algorithm outperforms the other two algorithms in recognition rate and stability.