Ship target automatic detection and identification method and system in natural scene

Abstract

The invention discloses a ship target automatic detection and recognition method in a natural scene, and the method comprises the following steps: collecting ship target video stream information, reading the ship target video stream information in a framing manner, and removing miscellaneous points in the ship target video stream information to generate a target image; performing data enhancementon the target image to generate an enhanced image and enrich data of the training data set; constructing a deep learning multi-scale feature fusion network, and transmitting the data of the training data set to the deep learning network for model training to generate a training model; detecting the ship target video stream in real time according to the training model, and automatically marking theship target in the ship target video stream. The method is suitable for multi-scene ship target automatic detection and identification, is high in detection precision under the weather conditions ofnight, rain, snow, heavy fog and the like, is good in detection effect on different types of ships with different sizes, effectively solves the problem of target missing detection, improves the detection accuracy, is high in anti-interference capability, is good in robustness, and can achieve the high-precision real-time detection of the ship in a complex environment.

Description

technical field [0001] The invention belongs to the field of image processing and pattern recognition, and in particular relates to a method and system for automatic detection and recognition of ship targets in natural scenes. Background technique [0002] With the development of domestic water transportation, the safety of water transportation has also received attention from all aspects. It is of great significance to accurately identify and detect the type and position of ships in real time for safe navigation of ships. In the process of ship detection, faced with problems such as the size of the ship target, the complex characteristics of the shore background, and weather interference, it is a great challenge to detect the ship target accurately in real time. [0003] The traditional target detection method first obtains the outline of the target object, and then classifies the image of the outline area. For real-time detected objects, the background difference method a...

AI Technical Summary

Problems solved by technology

Since the ripples on the water surface and the speed of the ship will interfere with the extraction of the target object area during actual detection, and the artificially designed feature extractor is less robust in the case of scale, rotation and noise, the model will be used in practice. In the detection process, the effect is poor and it is not universal

Images