BgCut: automatic ship detection from UAV images.

Area of Science:

• Computer Vision
• Image Processing
• Remote Sensing

Background:

• Ship detection in static Unmanned Aerial Vehicle (UAV) aerial images is crucial for maritime surveillance and precise positioning.
• Existing methods often require manual intervention or lack robustness in diverse natural sea conditions.
• Automated segmentation of sea targets from complex backgrounds remains a significant challenge.

Purpose of the Study:

• To develop an improved universal background model for automatic foreground object segmentation in UAV aerial imagery.
• To enhance the accuracy and efficiency of ship detection and precise positioning in maritime surveillance.
• To provide an adaptive and robust solution for automated industrial image processing.

Main Methods:

• An improved universal background model based on the Grabcut algorithm was proposed.
• A sea template library and region growing algorithm were used to generate an initial background trimap.
• The trimap initialized the Grabcut algorithm, enabling iterative-free segmentation of foreground objects.

Main Results:

• Extensive experiments on real UAV aerial images demonstrated the model's effectiveness.
• The proposed algorithm achieved adaptive and accurate sea target segmentation.
• The method showed good performance in precise positioning of detected ships.

Conclusions:

• The improved Grabcut-based model offers a robust and efficient solution for automatic ship detection in UAV imagery.
• The algorithm's adaptability and segmentation accuracy are suitable for maritime surveillance and related industrial applications.
• This approach facilitates automated processing of industrial images, advancing research in related fields.