In order to solve the problems of low classification performance, low statistical similarity and low mining accuracy of traditional data miningalgorithms, an incremental mining algorithm for sports video key pose dat...
详细信息
In order to solve the problems of low classification performance, low statistical similarity and low mining accuracy of traditional data miningalgorithms, an incremental mining algorithm for sports video key pose data based on depth learning is proposed. First, the training tag of depth learning is made by using analog signal matrix, and the implementation prospect of sports video key pose frame is extracted with Caffe (Convolutional Architecture for Fast Feature Embedding) open source framework. The interference region in key pose frame is removed by clustering algorithm, and the key pose region of sports video is obtained. Secondly, the SOFM (Self-Organizing Feature Map) network is used to cluster the data of the key pose area of sports video, and the incrementalmining model of the key pose data of sports video is established, and the data acquisition operation is carried out. The incrementalmining parameters of key pose data of sports video are obtained by using the combined paradigm, finally, the mining parameters are input into the mining model, and the incrementalmining of data is realized by using bwmorph method. The experimental results show that the key pose classification performance of the algorithm is much higher than that of the traditional sports video key pose data miningalgorithm, the statistical similarity is high, and the method has higher mining accuracy and is more suitable for the mining of the key gesture data of the sports video.
Compared with traditional distributed networks, the complex access environment, flexible access mode, massive access terminal, and data in an active distribution network will bring great security challenges to data tr...
详细信息
Compared with traditional distributed networks, the complex access environment, flexible access mode, massive access terminal, and data in an active distribution network will bring great security challenges to data transmission. The existing data security methods, such as access control and encryption, address the security of massive, high dimensional, and non-text data in the active distribution network. Therefore, feature selection algorithm based on rough set is first given to reduce the complexity of massive and high dimensional data. And then, based on feature selection, the authors propose a data filtering function model miningalgorithm by using gene expression programming (DFFM-FSGEP). Finally, to solve the data filter function model mining of the incremental dataset, they also present an incremental mining algorithm of the filtering function model based on functional fitting (IMFFM-FF). Experimental results show that the proposed algorithm in this study can greatly reduce the complexity of experimental datasets to be processed, and compared with the other algorithms, DFFM-FSGEP has higher classification accuracy and sensitivity, and IMFFM-FF has higher classification speed and classification accuracy for incremental datasets.
暂无评论