摘要
提出一种基于小波包多尺度信息熵的鱼类识别方法。该方法首先对鱼体的回波包络信号进行小波包分解,得到分布在不同频段内的分解信号,并提取各个频带内信号的信息熵作为识别特征量。对三种常见的不同形状的鱼类进行了水池试验,提取多尺度信息熵,并使用BP神经网络分类器成功进行了分类。结果表明:利用小波包多尺度信息熵作为特征量,可对不同形状的鱼类进行识别,且具有较高的识别率。
A method of fish identification based on wavelet packet multi -scale information entropy is proposed in this essay. Firstly, wavelet packet decomposition is done with the echo envelop signal of the fish, and the decomposed signals in different frequency bands are gotten. Then the information entropy of the signal in each frequency band is extracted as the identification characteristic feature. The pool experiments has been done with three kinds of common fish with different shapes, the multi - scale information entropy has been extracted, and the classification done by the BP neural network classifier is successful. The result reveals that the method of using wavelet packet multi - scale information entropy as the identification characteristic feature can identify fish of different shaps with high recognition properties.
出处
《网络新媒体技术》
2012年第4期47-52,共6页
Network New Media Technology
关键词
小波包
信息熵
特征提取
鱼类识别
wavelet packet, information entropy, feature extraction, fish identification
