A passive detection algorithm for low-altitude small target based on a wavelet neural network

A passive detection algorithm is presented for multiple low-altitude small targets, which employs a wavelet neural network (WNN). The slope, kurtosis, and skewness are employed as the features for low-altitude small target detection, and an algorithm is given to determine the number of targets. A WNN is used to establish a relationship between signal classes and the signal characteristics using training signals. Then, signals are classified as either target present or target not present using the WNN. Indoor data from a research laboratory and outdoor data from a bridge in the Jimo District, Qingdao, were used for training and evaluation. The performance results show that the error rate with the proposed WNN-based algorithm is better than those based on the slope, skewness, and kurtosis of signal. Furthermore, the proposed algorithm is better than those based on other neural networks such as BPNN, RBFNN, SOMNN, and SVM. At a distance of 3 km, the recognition rate is greater than 84%, which is better than other techniques such as visual recognition, acoustic, and active radar.