基于U-Net和Res_UNet模型的 传送带矿石图像分割方法

针对传送带矿石图像中矿石粘连和边缘模糊造成的分割不准确问题，提出了一种基于U-Net和Res_UNet模型的传送带矿石图像分割方法.该方法首先将待分割图像经过灰度化、中值滤波和自适应直方图均衡化处理后，利用预训练的U-Net模型提取图像轮廓;然后，将图像轮廓二值化后，利用预训练的Res_UNet模型进行轮廓优化;最后，利用OpenCV得到分割结果.与基于形态学重建的分水岭算法和NUR法分别对10张测试图进行实验比较，结果表明，所提出的利用深度学习实现矿石轮廓检测和优化方法分割的结果更加准确，证明了其对传送带矿石图像分割的有效性.

Ore Image Segmentation Method of Conveyor Belt Based on U-Net and Res_UNet Models

Aiming at the problem of inaccurate segmentation caused by the adhesion and edge blurring of the ore image in the conveyor belt， a method for ore image segmentation of conveyor belt based on U-Net and Res_UNet models is proposed. Firstly， the image to be segmented is processed by gray-scale， median filtering and adaptive histogram equalization， and then the pre-trained U-Net model is used to extract the image contour. Then， after binary image contour， the pre-trained Res_UNet model is used for contour optimization. Finally， OpenCV is used to obtain the segmentation result. Compared with watershed algorithm based on morphological reconstruction and NUR method for 10 test images， the proposed method for ore contour detection and optimization based on deep learning is more accurate， proving its effectiveness for image segmentation of conveyor belt ores.