Research On Detection Algorithm For Underwater Objects Based On Deep Learning

In recent years,deep learning has gained tremendous development,and the generic object detection algorithms based on deep learning are becoming mature and applied in industrial production environments.Generic object detection is an important direction of deep learning.This type of algorithm can analyze and detect the position and bounding box information of objects through the input image.However,when this technology is applied to underwater scenes,its detection accuracy will decrease,because underwater images have image degradation,which is mainly caused by the absorption of light by the water body and the scattering of light by floating particles.In order to reduce the influence of underwater image degradation on the object detection algorithm,the original image is usually preprocessed to improve the clarity of the image,which makes the algorithm easier to identify the object.Image enhancement methods mainly include traditional methods and neural network methods,but a certain method is usually only applicable to a specific scene.This article optimizes the object detection algorithm for the underwater image degradation,combines the image enhancement algorithm with the generic object detection algorithm YOLO,and utilizes the image fusion method and the WEIoU bounding box regression loss function to improve the underwater object detection algorithm accuracy.The image fusion method fuses the original image with the enhanced image,and the algorithm can select the appropriate feature map according to the actual effect of the enhanced image,which is conducive to improving the robustness of the algorithm.The bounding box loss function chooses the WEIoU loss function,which is not sensitive to the scale of the bounding box,and describes the position and shape information between the bounding boxes more accurately,and assigns different weights to different object categories,which can improve the accuracy of underwater object detection algorithm.Through our experiments,it can be seen that the optimization method proposed in this article improves the accuracy of underwater object detection algorithm to a certain extent.