A Semantic Segmentation Method for Panoramic Water Image

Abstract

The invention discloses a method for semantic segmentation of water surface panoramic images, comprising: designing a convolutional neural network, performing pixel-level marking on each target category in a water surface panoramic image in a training set to obtain a real marked image, using the training set to train the convolutional neural network, Compare the output of the convolutional neural network with the real labeled image to obtain the training error, backpropagate the training error in the convolutional neural network, update the parameters of the convolutional neural network, and get the trained one after multiple iterations of training Convolutional neural network. The water surface panoramic image to be tested is input into the convolutional neural network for real-time semantic segmentation, and the segmentation result of the water surface panoramic image is obtained. The invention has fast segmentation speed and good segmentation effect on small area targets. Provide comprehensive, fast and accurate environmental perception information for surface smart devices such as unmanned boats.

Description

technical field [0001] The invention belongs to the technical field of computer vision recognition, and more specifically relates to a method for semantic segmentation of water surface panoramic images. Background technique [0002] With the proposal of the ocean power strategy, my country has begun to vigorously develop marine equipment. As an unmanned surface mobile platform, unmanned boats can not only enter some harsh environments to complete tasks, but also improve the efficiency of tasks that require long-term operations. They play a very important role in many practical applications, such as Customs patrols, shallow sea mine clearance and water quality monitoring, etc. For unmanned boats, environmental perception technology is an indispensable part of autonomous navigation and autonomous obstacle avoidance. In particular, in the process of actual operation, it is far from enough to only be able to detect targets or obstacles in the direction of advancement, and dang...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the above defects or improvement needs of the prior art, the present invention provides a method for semantic segmentation of panoramic water surface images, thereby solving the technical problems existing in the prior art that the speed is slow and the segmentation effect on small area objects is not good

Images