A coal-gangue optimization identification method
-
摘要: 针对输送带磨损造成煤矸石图像目标检测不准确、影响煤矸石识别准确率等问题,提出了一种煤矸石优化识别方法。采集的图像经过裁切、去噪、灰度化等预处理后,利用训练好的CornerNet-Squeeze深度学习模型判断图像中是否存在待检测的煤或矸石,若存在则定位煤或矸石在图像中的位置,有效降低目标检测时输送带背景干扰;对定位区域进行灰度直方图分析,依据图像灰度直方图的三阶矩特征参数对煤矸石进行分类,判定是煤还是矸石,提高识别准确率。实验结果表明,该方法识别准确率为91.3%,单张图像识别时间为41 ms,具有较高的识别准确率和较好的实时性。Abstract: Aiming at problem that target detection of coal-gangue image is not accurate due to wear of conveyor belt, which affects identification accuracy of coal-gangue, a coal-gangue optimization identification method is proposed. After pre-processing of collected images such as cutting, denoising and grayscale, the trained cornernet-squeeze deep learning model is used to judge whether there is coal or gangue to be detected in the images. If there is, position of coal or gangue in the images is located, which can effectively reduce background interference of conveyor belt during detection. The location area is analyzed by gray histogram, then according to third moment characteristic parameter of image gray histogram, coal-gangue is classified to determine whether it is coal or gangue to improve identification accuracy. The experimental results show that the method has high identification accuracy and good real-time performance with identification accuracy of 91.3% and identification time of 41 ms for single image.
点击查看大图
计量
- 文章访问数: 105
- HTML全文浏览量: 12
- PDF下载量: 12
- 被引次数: 0