欢迎访问林业科学,今天是

林业科学 ›› 2020, Vol. 56 ›› Issue (12): 91-100.doi: 10.11707/j.1001-7488.20201211

• 论文与研究报告 • 上一篇    下一篇

3种天牛幼虫蛀食振动小波包特征及其识别

刘圣煌1,杨江天1,崔建新2,*   

  1. 1. 北京交通大学机械与电子控制工程学院 北京 100044
    2. 河南科技学院害虫天敌繁育研究中心 新乡 453003
  • 收稿日期:2019-05-22 出版日期:2020-12-25 发布日期:2021-01-22
  • 通讯作者: 崔建新
  • 基金资助:
    河南省科技开放合作项目(172106000056);国家自然科学基金项目(31772501)

Characteristics and Identification of Wavelet Packet of Wood Boring Vibration by Three Longicorn Larvae

Shenghuang Liu1,Jiangtian Yang1,Jianxin Cui2,*   

  1. 1. School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University Beijing 100044
    2. Breeding Center of Natural Insect Enemies for Pests, Henan Institute of Science and Technology Xinxiang 453003
  • Received:2019-05-22 Online:2020-12-25 Published:2021-01-22
  • Contact: Jianxin Cui

摘要:

目的: 蛀干害虫幼虫生活在树干内部,种类鉴别困难。为了迅速准确地判断天牛幼虫种类,进行科学防治,试验通过检测分析天牛幼虫蛀食树干振动信号提取害虫行为学特征,以鉴定天牛种类。方法: 获取天牛幼虫在蛀食通道内活动时发出的振动信号,分析信号波形和相对幅值,选择天牛幼虫蛀食树干振动信号,抽取振动信号波形和持续时间等时域特征。首先用变分模态分解对振动信号做降噪处理,再对降噪后的信号进行小波包3层分解,计算各节点的频域能量,以特征频率和节点能量占比作为频域特征。时域、频域特征相结合识别天牛种类。结果: 云斑天牛蛀食振动持续时间约为11.1 ms,衰减震荡波形中部存在一个明显的下凹;特征频率约为550 Hz,小波包能量主要集中在第2、3、4节点。光肩星天牛蛀食振动持续时长约为8.2 ms,衰减震荡波形在中部和尾部各有一个下凹;与云斑天牛相比,蛀食振动信号频谱无550 Hz频率成分,但存在2 500 Hz频率分量。桃红颈天牛(为害紫叶李)蛀食振动持续时间约为6.4 ms,波形无明显下凹,频谱图上有2个2 000 Hz以上的频率成分;桃红颈天牛(为害西府海棠)蛀食振动持续时间约19.8 ms,波形无下凹,在尾部有上凸,频谱存在100 Hz的低频频率成分。结论: 采用天牛幼虫蛀干振动信号分析法能有效识别天牛种类,对实现天牛幼虫活体快速林间检测,制定科学防治方案有重要实用价值。

关键词: 天牛, 幼虫, 蛀木振动, 小波包分解, 变分模态分解

Abstract:

Objective: Since the larvae of forest trunk borers live inside the trunk, it is difficult to identify the species. In order to quickly and accurately recognize the species of longicorn larvae and carry out biological control, a novel method was proposed to detect and analyze the vibration signals of longicorn larvae feeding in trees. With the method, the behavioral characteristics extraction and species identification of longicorn larvae were performed by means of wood boring vibration signal processing. Method: The vibration signals were obtained when the longicorn larvae were moving in the feeding channel. The signal waveform and relative amplitude were analyzed, then the time domain characteristics, such as waveform and vibration duration were extracted. Firstly, variational mode decomposition (VMD) was used for signal denoising, and then the denoised signal was decomposed by wavelet packet. The energy of each node in frequency domain was calculated. The characteristic frequency and energy ratio of each node were employed as the frequency domain characteristics. The time-domain and frequency-domain features were used to identify the longicorn species. Result: The duration of Batocera horsfieldi boring wood vibration signal was about 11.1 ms, and there was a distinct concave in middle of the waveform. The characteristic frequency was about 550 Hz, and the energy of wavelet packet was mainly distributed at 2, 3, 4 nodes, which is different from that of Anoplophora glabripennis. The duration of Anoplophora glabripennis boring wood vibration signal was about 8.2 ms. There were two concaves in the middle and real of the waveform separately. Compared with those of Batocera horsfieldi, there was no 550 Hz frequency component, but a 2 500 Hz component in the spectrum of A. glabripennis boring wood vibration signal. The duration time of Aromia bungii (in Prunus cerasifera) boring wood vibration was about 6.4 ms without concave in the waveform. Two frequency components above 2 000 Hz were found in spectrum. The vibration duration of A. bungii (in Malus micromalus) boring wood was about 19.8 ms. There was no concave but a convex in the rear of the waveform. Low frequency component near 100 Hz was found in the spectrum. Conclusion: The longicorn larvae can be detected and identified by wood boring vibration signal analysis. It is of great importance to early recognize trunk borer and select appropriate control measures.

Key words: longhorn, larva, wood boring vibration signal, wavelet packet decomposition, variational mode decomposition

中图分类号: