Large-format remote sensing image ship target detection method and system under small sample condition

Abstract

The invention discloses a large-format remote sensing image ship target detection method and system under a small sample condition. The method comprises the steps: introducing an SENet attention mechanism module into a Bottleneck module of a YOLO v5 network, adding a detection layer, and forming a target detection network; constructing a pre-training data set to perform pre-training on the targetdetection network, and performing transfer learning on the target detection network by using the preprocessed ship target annotation data set to obtain a test model; and carrying out iterative cuttingon the large-format remote sensing image to be recognized to obtain a small image, respectively carrying out target detection and target coordinate conversion by using the test model, and outputtinga target detection result. According to the method, effective training of a network model and large-format image quick detection can be completed by using small-batch image target samples, and the accuracy and robustness of ship target detection are maintained.

Description

technical field [0001] The invention relates to the technical field of image target recognition, in particular to a large-format remote sensing image ship target detection method and system under the condition of small samples. Background technique [0002] Ship target detection plays a prominent role in the fields of national maritime security, monitoring illegal fishing, and combating illegal smuggling. In the past few years, with the rapid growth of data volume and continuous improvement of computing power, deep learning has made a series of breakthroughs in the field of target detection. For example, the YOLO series of target detection algorithms have been continuously developed and applied. The latest version is YOLO v5 has been released https: / / github.com / ultralytics / yolov5, and its network structure is mainly composed of Focus, Conv, Bottleneck, BottleneckCSP, Upsampling, concat, CSP, SPP, Conv2d modules. However, applying deep learning algorithms such as YOLO direct...

AI Technical Summary

Problems solved by technology

In the field of remote sensing images, constructing a data set that can cover the complete sample distribution often requires a lot of manpower and material resources to collect and label data, and the training set and test set come from the same domain. When encountering a new target, the model The detection performance will be greatly reduced, and there will be an overfitting problem.

Images