Multi-scene sailor unsafe behavior detection method based on improved YOLOv3

Abstract

The invention provides a multi-scene sailor unsafe behavior detection method based on improved YOLOv3. Water traffic accidents directly or indirectly caused by human factors account for about 80% of the total number of the water traffic accidents. According to the invention, the improved YOLOv3 network is utilized to detect six types of unsafe behaviors of sailors in different scenes, such as off-duty, smoking, mobile phone playing, conversation, sleeping and lookout negligence, and the detected unsafe behaviors are calibrated and reminded. The method comprises the following steps: firstly, adding a CLAHE algorithm into a data enhancement part; secondly, changing the original three-scale detection into four-scale detection, and introducing a negative feedback mechanism into a feature fusion network part, so the features of all scales are better fused; and finally, improving a non-maximum suppression algorithm to carry out Gaussian weighting on a candidate frame to obtain a more accurate result. According to the invention, unsafe behaviors of the sailors can be accurately detected and the sailors can be reminded, so the occurrence rate of marine accidents caused by human factors is reduced and the safe navigation capability of a ship is improved.

Description

technical field [0001] The invention relates to the fields of computer vision and shipping safety, in particular to a multi-scenario crew unsafe behavior detection method based on improved YOLOv3. [0002] technical background [0003] The special occupation of the crew, the psychological quality of the crew, the harsh environment of the voyage, the negligence of the crew training, the low quality of the crew, and the lack of experience will all cause unsafe behaviors during navigation. At sea, the unsafe behavior of the crew may cause casualties, property damage, environmental pollution, etc. Statistics show that in water transport accidents, the accidents directly or indirectly caused by human factors account for about 80%. [0004] The crew's unsafe behavior refers to the seafarer's wrongful behavior that violates laws and regulations or safe operation rules or regulations under the domination of his own consciousness, making accidents possible or likely to occur, thus en...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to effectively detect and identify six types of unsafe behaviors of crew members on the ship's bridge, such as leaving the post, smoking, playing with mobile phones, talking, sleeping, and watching and negligence, and timely alarming, so as to solve the problem of detection and identification of unsafe behaviors of crew members during navigation , improve the overall safe navigation capability of the ship

Images