A Target Recognition System for Small Sample Underwater Images

Abstract

The invention discloses a target recognition system for small sample underwater images. The results are evaluated; the method and system integrates a very comprehensive image processing operation and deep learning model, which can help users to use different ways and different combinations to achieve underwater target training and recognition, and compare the prediction effect. The combined scheme is for reference in subsequent research, effectively solving the problem that in the target recognition research of small sample underwater images, there is no scheme to follow, no method to follow, and is limited by personal professional experience, and the present invention can achieve deep The automatic deployment and accelerated learning of the learning framework can be deployed on multiple development platforms, which is easy to promote and use, and greatly saves manpower and time costs.

Description

technical field [0001] The invention relates to the technical field of underwater target recognition, in particular to a target recognition system for small-sample underwater images. Background technique [0002] In underwater target recognition, due to the limitations of the environment, optical imaging is blocked, and acoustic imaging can only be used in long-distance directions, that is, target search is carried out with the help of sonar equipment. However, due to the changeable underwater acoustic medium, the signal transmission process is susceptible to interference, resulting in low quality sonar images of targets, high noise, and indistinct features. [0003] Due to the unknown of the underwater environment, most of the targets are small samples and rare, so the acquired target image data is very little. When processing such target images, we can only rely on the experience of researchers to operate without mature processing There is no systematic model to train und...

AI Technical Summary

Problems solved by technology

However, due to the changeable underwater acoustic medium, the signal transmission process is susceptible to interference, resulting in low quality sonar images of targets, high noise, and indistinct features

[0003] Due to the unknown of the underwater environment, most of the targets are small samples and rare, so the acquired target image data is very little. When processing such target images, we can only rely on the experience of researchers to operate without mature processing There is no systematic model to train underwater images in a targeted manner. When analyzing underwater targets, the image processing process is complicated and the correlation between modules is not strong. There is no mature processing method for unknown and small-sample underwater images. For reference, a lot of human resources and time resources are wasted. For these small samples and unfamiliar data, the image processing method can only be selected tentatively, and the verification is not enough, which seriously affects the target recognition efficiency.

Images