Constant false alarm rate detection of slow targets in polarimetric along‐track interferometric synthetic aperture radar imagery

Automatic deep learning for automatic target detection of sonar images

“…, which are shown in Table 1. The advantages of using two different customized genotypes are mainly in two aspects: one is that it can enhance the attention of the two types of convolutional cells during architecture search, which also helps to shorten the search duration; the other is to reduce the GPU memory requirements during the search period, compared with the standard darts, comparing to the shared (8) O used in the standard DARTS, our customized genotypes saves three times (8/6+8/5) less memory.…”

“…Imaging sonars [1][2][3] are indispensable sensors in underwater remote sensing fields, which can provide abundant visual information of the observed sea floor. Many studies [2][3][4][5][6][7] in recent years has focused on the automatic target detection of sonar images, which have a wide range of applications, including mine detection, underwater target search, and recovery, et al Traditional detectors such as Generalized Likelihood Ratio Test (GLRT) [8] and Constant False Alarm Rate (CFAR) [8] need to manually design statistical threshold for different distributions, yet not only bad stability and poor performance are quite common when applying traditional detectors to capture targets from actually measured data, but they cannot identify targets either without using additional recognition algorithms. As a result, in recent years, deep learning-based target detection of sonar images has started booming owing to the conciseness, high performance, and the ability to detect and classify targets simultaneously.…”