像素重要性测量特征下的侧扫声呐目标检测

研究了在较低信噪比下,在保证检测概率的前提下尽量降低虚警概率的目标检测,提出了一种针对特定目标的两阶段筛选算法.第一阶段中,首先使用阈值分割出有效点,并定义了一种新的像素重要性测量特征用于初步筛选目标。即通过有效像素点之间的距离来赋以高斯分布的权值,当前像素重要性的值定义为剩余有效点的距离加权和,具有较高的像素重要性值的聚集性强的区域内像素点会被定位出来。第二阶段,使用卷积神经网络分类器排除虚假目标.在实验中,使用近期无人潜器获得的海底数据,召回率与虚警概率分别达到90.39%与2.39%,证明了其相比声呐目标检测主流算法有更好的检测能力。

Target detection algorithm in side-scan sonar image based on pixel importance measurement

The prior task is to stabilize the detection ability and reduce false alarm probability simultaneously underlow signal-noise ratio condition. A two-stage novel pixel importance measurement algorithm in the side-scan sonar detection application is proposed. In first stage of the algorithm, a new feature defined as Pixel Importance Value(PIV) is proposed based on distances between the target pixel and each other pixels. PIV measurement of current pixel is defined as the weighted sum of all remaining segmented pixels. The weighted part refers to Gaussian kernel, which means closer pixels gets higher weight. Thus, targets with higher pixel importance value can be located. In the second stage, we use convolutional neural network as classifier to eliminate the dot-like false targets in terms of targets shape. In our experiments based on recent underwater data obtained by autonomous underwater vehicle, we demonstrate superior performance of our algorithm over the state-of-the-art sonar detection algorithms in terms of 90.39% recall rate and 2.39% false alarm probability.