Radar RD image target detection method under low signal-to-noise ratio based on deep learning

Abstract

The invention discloses a radar RD image target detection method under a low signal-to-noise ratio based on deep learning. The method comprises the following steps: obtaining radar echo data, and generating a radar RD image; preprocessing and labeling the radar RD image to obtain data sets; classifying the data sets to obtain a training set, a verification set and a test set; constructing a deep learning neural network for target detection under a low signal-to-noise ratio; training the constructed neural network by adopting the training set, and outputting a loss value and a trained detection neural network; performing target detection on the test set by using the trained detection neural network; and obtaining a test set target detection accuracy result. According to the invention, the optimal radar target detection network is obtained through training by the deep learning neural network based on a large amount of radar RD image data, the target detection network obtained by the method can perform effective target detection on a radar RD image under a low signal-to-noise ratio, and has the advantages of high accuracy, good practical effect and the like.

Description

technical field [0001] The invention relates to the technical field of target detection, in particular to a radar range-Doppler (RD) image target detection method under low signal-to-noise ratio based on deep learning. Background technique [0002] In recent years, object detection has been widely used in many fields. As an important means of target detection, radar can analyze and process the echoes of the illuminated area, detect target information from signals such as clutter, interference, noise, etc., and determine its distance, speed, angle and other parameters. However, in complex backgrounds, especially with low SNR, the echo information often contains a large amount of noise and clutter information in addition to the target, and the noise is close to the target amplitude, and the clutter often presents nonlinear, non-Gaussian, non-linear Uniform and non-stationary characteristics, which greatly limit the radar target detection performance. [0003] Existing radar ...

AI Technical Summary

Problems solved by technology

However, in complex backgrounds, especially with low signal-to-noise ratios, the echo information often contains a large amount of noise and clutter information in addition to the target, and the amplitude of the noise is close to the target, and the clutter often presents nonlinear, non-Gaussian, and non-linear. Uniform, non-stationary nature, which greatly limits radar target detection performance

[0003] Existing radar target detection methods include constant false alarm detection, machine learning algorithms, etc. CFAR algorithms are based on statistical models, and it is often difficult to accurately describe the background model. Under the signal-to-noise ratio, there will be serious constant false alarm loss, and the detection performance will decline; machine learning algorithms such as self-correcting extreme learning machine and other target detection algorithms are difficult to extract target features in depth, and the ability to distinguish between targets and backgrounds is weak, and it is difficult to achieve accurate target classification , which greatly increases the difficulty of target detection and limits the detection performance

In summary, the existing radar RD image target detection methods have problems such as simple model, low universality, and weak learning ability, and it is difficult to fundamentally solve the problem of weak radar RD image target detection ability under low signal-to-noise ratio.

Images