A Small Target Detection Method on Sea Surface Based on Markov Transition Field

In response to the limitations of manually extracting features from target signals in the background of sea clutter, this paper studies Markov transition fields and convolutional neural networks, and proposes a sea small target detection method based on Markov transition fields, which transforms the detection problem into a binary classification problem. The one-dimensional observation echo is transformed into a two-dimensional image through the Markov transition field, and a sea clutter and weak signal classification model is established to complete the task of detecting weak signals in sea clutter. Taking IPIX measured radar data as the experimental object, a transfer learning model is built to autonomously learn MTF image features, improve the performance of the ResNet model, reduce training costs, compare the effects of different sampling points on detection results, and find that the effect is optimal when the sampling point is 1024. Compared with other image coding methods and classification networks, the proposed method can deeply explore the differences between targets and clutter, and has better detection and classification performance.