Underwater polarization image restoration and identification method

Abstract

The invention provides an underwater polarization image restoration and identification method. The method comprises the following steps: constructing an underwater active imaging system; shooting a polarization image; constructing a data set; carrying out marking processing on the data set; designing an image restoration network and an identification network on the basis of the pytorch framework; constructing an image restoration and target identification network layer; and combining the restored image and the recognition result to realize the restoration and recognition of the underwater polarization image through a plurality of steps such as end-to-end image restoration and recognition. And finally, testing the quality effect of the image restored by the image restoration network model by using the test set, and testing the recognition effect of the recognition network by using the label file containing the target position and category information of the test set.

Description

technical field [0001] The invention relates to the technical field of underwater optical imaging, in particular to an underwater polarization image restoration and recognition method. Background technique [0002] Underwater optical imaging technology is widely used in underwater exploration and detection, and provides technical support for exploring the ocean. However, traditional underwater optical imaging technology cannot filter and eliminate the scattered light emitted by impurity particles, resulting in poor imaging quality. . Therefore, the research on the combination of optical physical model and deep learning is proposed, that is, the neural network is used to extract the effective polarization information of the image to achieve the purpose of clear imaging. However, there are still deficiencies and limitations in this research for realizing the intelligence of underwater detection equipment. [0003] At present, image restoration techniques that process image p...

AI Technical Summary

Problems solved by technology

[0004] For example, the patent No. CN113538279A discloses a deep learning-based underwater polarization image restoration system and its restoration method, which restores high-turbidity underwater images by using convolutional neural networks, although the underwater polarization image restoration system of this patent is based on depth Learning, only processing the polarization information that can improve the imaging quality, using the extracted polarization information to remove the influence of the scattered light of the turbid particles to achieve clear imaging can restore the image of the underwater target, but ignores the polarization information The characteristic information of the implied image content

The lack of processing of the characteristic information of the hidden image content makes the system unable to respond and process the image content accordingly

Images