自适应分段时域质心特征在鱼类识别中的应用

Applications of adaptive segmentation temporal centroid features in fish identifications

提出了一种基于时域质心的时域自适应分段方法。该方法以时域质心为依据对信号的时域进行划分,在划分的各个子段内计算时域质心,并将其作为下一层划分的分割点。各个子段内的时域质心反映了信号的能量分布特性,可作为识别特征量。对三种常见的不同形状的鱼类进行了水池试验,提取自适应分段时域质心特征,并使用BP神经网络分类器成功进行了分类。结果表明:利用自适应分段时域质心特征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 the adaptive segmentation based on the temporal centroid is proposed in this paper This method divides signals in the time-domain according to the temporal centroid, and then calculates the temporal centroid in the divided sub-segmentation, and makes the temporal centroid to be the split point of the next layer. The temporal centroid in each sub-segmentation indicates the energy distribution characteristics, which can act as the identification features. The pool experiments of three kinds of fish with different shapes have been conducted, sub-segment temporal centroid characteristics have been extracted, and the targets have been classified by applying the BP（Back Propagation） neural network classifier successfully. The results suggest that fish with different shapes can be recognized using their adaptive segment temporal centroid features and the recognition rate is high.