首页|多特征反向传播-人工神经网络微钻阻力年轮识别方法

多特征反向传播-人工神经网络微钻阻力年轮识别方法

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
峰谷年轮识别算法仅使用峰谷差值这一个特征进行年轮识别,因此该算法的误判率和漏判率较高.为了进一步提高微钻阻力年轮识别精度,提出了一种基于多个波峰特征的反向传播-人工神经网络(BP-ANN)年轮识别方法.首先使用峰谷年轮算法识别有效波峰,然后使用波峰阻力值、波峰与前波谷和后波谷的阻力差值、波峰与前波谷和后波谷的距离、前波谷与后波谷的距离等6个参数描述波峰特征;然后根据阻力图与圆盘图像确定有效波峰的类型,如果该波峰是一个年轮信号,则标记为"1",否则标记为"0";最后使用BP-ANN算法构建有效波峰分类模型.结果显示,BP-ANN模型的准确率比峰谷年轮识别算法提高了1.26个百分点,误判率和漏判率比峰谷年轮识别算法分别减少了1.06和1.38个百分点.结果表明:基于多个波峰特征的BP-ANN模型的年轮识别方法可行;与传统的峰谷年轮识别算法相比,该方法可有效提高年轮识别精度,有效降低年轮误判率和漏判率.
A Method for Identifying Annual Rings with Multiple Features in Micro Drilling Resistance Based on Back Propagation-Artificial Neural Network
The peak-valley annual ring recognition algorithm only uses the feature of the difference between the peak value and valley value for annual ring recognition,so the algorithm has a high rate of false positives and false negatives.To enhance the accuracy of micro-drill resistance ring recognition,the Back Propagation-Artificial Neural Network(BP-ANN)algorithm was employed.Firstly,the peak-valley ring recognition algorithm was utilized to identify the effective peaks.Secondly,characteristics such as peak resistance value,resistance difference between the peak and adjacent valleys,distance between the peak and adjacent valleys,and distance between front and back valleys were employed to describe these peaks.Based on the analysis of resistance graphs and disk images,the effective peaks were classified accordingly.If the peak was an annual ring signal,it was marked as"1";otherwise,it was marked as"0".Finally,an effective peak classification model was constructed using BP-ANN algorithm.Compared to the peak valley annual ring recognition algorithm,the BP-ANN model improved the accuracy by 1.26 percentage points and reduced false positives and false negatives by 1.06 and 1.38 percentage points,respectively.The results indicated that the BP-ANN model based on multiple peak features was feasible for identifying annual rings.Compared with the traditional peak-valley annual ring recognition algorithms,the proposed method can effectively improve the accuracy of annual ring recognition and reduce the misjudgment and omission rates of annual rings.

Back Propagation-Artificial Neural Network(BP-ANN)micro-drill resistance instrumentpeak-valley annual ring recognition algorithmannual ring

姚建峰、吴振洋、胡雪凡、孙艳歌、田文静、路一曼、李晓

展开 >

信阳师范大学 计算机与信息技术学院,河南 信阳 464000

北京市园林绿化科学研究院 生态修复研究所,北京 100102

反向传播-人工神经网络(BP-ANN) 微钻阻力仪 峰谷年轮识别算法 年轮

国家自然科学基金项目河南省自然科学基金项目河南省高等学校重点科研项目河南省研究生质量工程项目

3207176123230042116722A220002YJS2023SZ23

2024

10.3969/j.issn.1003-0972.2024.04.007

信阳师范学院学报(自然科学版)
信阳师范学院

信阳师范学院学报(自然科学版)

CSTPCD
影响因子：0.446
ISSN：1003-0972
年,卷(期)：2024.37(4)