1.School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China 2.Key Laboratory of Information Photonics Technology, Ministry of Industry and Information Technology, Beijing Institute of Technology, Beijing 100081, China
Fund Project:Project supported by the National Natural Science Foundation of China (Grant No. 61835001)
Received Date:25 March 2021
Accepted Date:07 April 2021
Available Online:07 June 2021
Published Online:20 September 2021
Abstract:This paper proposes an underwater ranging method based on wavelet transform. First, according to the band-pass filtering characteristics of the wavelet transform, the time-domain signal is decomposed in the frequency domain. The wavelet basis functions with high similarity are established. These wavelet basis functions contain complete frequency domain information of time-domain signals. This method can improve the ability to decompose frequency domain of time-domain signals and extract the information about the effective frequency domain. Then, using the multiple frequency domain decomposition approximations, the effective frequency domain information contained in the time domain signal is completely extracted.The time-frequency signal of wavelet time-frequency fusion ranging takes the energy consistency of the time-frequency domain signal as the link and uses the binary spline interpolation structure to realize the time-frequency combination of the signal. In this method, the time-domain signal is first decomposed and filtered by wavelet time-domain to obtain more complete time-domain effective information. But at this time, the time-domain signal is the superimposed form of frequency-domain information, so the energy domain information contained in the time-frequency signal is decomposed into the wavelet frequency domain through the binary spline interpolation, and the energy expression form of the time-frequency signal can be obtained. The target is locked by finding the position of the maximum value of energy corresponding to the time-frequency domain of the signal to achieve the purpose of precise ranging. By performing the wavelet multi-layer time-domain decomposition filtering first, the frequency domain decomposition range can be effectively reduced, thereby avoiding data redundancy and reducing the ability to realize the effective frequency domain resolution.By using this method we successively carry out continuous light underwater ranging experiments with different attenuation length water bodies and different modulation frequencies, and analyze the influence of this method on continuous light underwater detection. Experiments verify that this ranging method successfully achieves the accurate measurement of targets within 8 attenuation lengths within an output power of 2.3 W, and its ranging accuracy is less than 1 cm; the use of wavelet time-frequency fusion ranging can pass the frequency domain energy decomposition capability enhancement, to a certain extent, compensates for the measurement error caused by the significant attenuation of the effective signal. Therefore, the ranging method can be applied to signals with complex frequency domain information or including a bandwidth. Keywords:wavelet transform/ underwater ranging/ band-pass filtering/ time-frequency combination
表1不同衰减系数与距离功率的关系 Table1.Relationship between different attenuation coefficients and distance power.
24.2.基于小波变换的时频分析法水下目标测量 -->
4.2.基于小波变换的时频分析法水下目标测量
采用PIN探测器接收的不同位置回波信号和同频调制的参考信号波形, 锁相放大器输出相应信号波形包含的相位信息和调制频率关系, 也可直接使用数据采集卡采集与输出此类信息. 衰减系数为0.99 m–1的水体, 调制频率50 MHz为起始频率, 记录50—200 MHz锁相放大器求解出的相位关系, 如图3(a)所示. 图3(b)为图3(a)的频率相位关系进行逆傅里叶变换. 通过图3(b)可知时域幅度曲线的半高宽范围, 可得出有效信息的频域范围约为130—180 MHz, 对该时间幅度信号进行小波频域分解, 获得信号时频空间的能量极值位置. 根据图3(c)可得基准位置能量极值对应时间为0.19646 μs, 1 m目标位置能量基准对应时间为0.20545 μs. 室温下, 海水折射率约为1.339[21], 则对应基准与目标距离为1.0071 m, 测量误差为0.71 cm, 该测量误差为精度0.1 cm的皮尺标定距离与测量距离之差. 图 3 回波信号与参考信号波形及相应运算结果 (a) 频率-相位差波形图; (b) 图(a)的傅里叶逆运算结果; (c), (d) 基准位置和1 m目标位置的时频能量极值的时域表示 Figure3. Echo signal and reference signal waveforms and corresponding calculation results: (a) Frequency vs. phase difference waveform diagram; (b) the Fourier inverse calculation result of panel (a); (c) and (d) time-frequency energy extreme value positions of the reference position and 1 m position of the target in the time-domain.
水体衰减系数为0.99 m–1, 分别把目标放置在1.5和2.5 m的位置, 将锁相放大器输出的数据进行记录或直接用数据采集卡采集相位关系, 将变换后的时域信号的幅值进行频域谱分解, 最小频率间隔分别为0.1, 0.075, 0.05和0.025 MHz. 图4(a)和图4(b)为同一水体和目标条件下, 不同频率间隔的测距结果. 图4(c)和图4(d)为不同衰减程度的水体和目标条件下, 不同频率间隔的测距结果. 由图4(a)—(d)的测距结果可以得到, 随着频域分解间隔的减小, 目标测量误差减小. 图 4 不同频率分解间隔对探测目标的影响 (a), (b) 同一水体条件下, 不同频率分解间隔时1.5和2.5 m目标的测量误差; (c), (d) 不同衰减系数的水体条件下, 频域分解间隔为0.1和0.025 MHz时不同目标距离的测量误差 Figure4. Influences of different frequency decomposition intervals on detection targets: (a), (b) Measurement error at differently frequency decomposition intervals for 1.5 and 2.5 m under the same water body condition; (c), (d) under the water bodies with different attenuation coefficient conditions, the measurement error different distance targets at frequency domain decomposition interval of 0.1 and 0.025 MHz, .
探究不同衰减长度水体下, 不同调制频率区间对该测距方法测量结果的影响. 如图5所示, 将发射源调制频率区间调节至50—200 MHz和700—850 MHz, 采用小波变换时频融合测距法对不同衰减长度水体的目标进行测量. 图5(e)和图5(f)是衰减长度为1—8的水体、调制频率区间为50—200 MHz和700—850 MHz时, 小波时频融合测距方法对目标的测量误差, 以衰减长度起伏表示测量误差波动范围, 得到两调制频率区间内实际探测距离与标定的目标距离的差值绝对值. 通过对图5(e)和图5(f)两组数据的测量误差进行比较, 可以发现低频调制频率范围的测量误差明显小于高频调制频率范围. 图5(a)—(d)给出衰减长度为4.5和6的水体下, 调制频率区间为50—200 MHz和700—850 MHz时通过小波时频变换得到的目标频域极值的时域能量表示, 可以明显地看出低频调制的时域能量极值位置波动范围小于高频调制时域能量极值位置波动范围. 下面从两方面分析低频调制与高频调制对该测距方法的影响: 1) 高频调制对后向散射的抑制作用优于低频调制, 经过高频调制后整体有效信号强度明显提升; 消除大量散射信号的影响, 有效信号波动范围明显降低. 2) 信号源为连续单模激光器时, 回波信号的有效频域波动范围极小. 基于小波变换的时频融合测距依赖于时频域的带通滤波特性, 即信号应包含丰富的频域信息. 低频调制的回波信号包含较多的杂散光频域信息. 这些杂波信号频域分布范围广阔, 有效信号频域分布集中, 能量强度较大, 易于分析提取; 高频调制频率的回波信号中, 杂散光明显减少, 可以获得较为完整的有效信号频域信息. 同时, 由于采样点数与时间间隔限制其频域分解能力, 可知其在固定频域间隔下的小波时频能量分解的频域分布较为分散, 易出现极值位置不明显或多极值现象, 导致高频调制作用的小波时频融合测距的测量误差以及各组测量误差波动范围明显增加. 图 5 调制频率(50—200 MHz和700—850 MHz)对小波时频融合测距方法的影响 (a)—(d)在4.5和6个衰减长度下, 探测目标在极值频率位置的时域能量表示; (e), (f) 两频率范围下不同衰减长度对应的测量误差 Figure5. Influences of modulation frequency on wavelet time-frequency fusion ranging method, where the modulation frequency is a range of 50 to 200 MHz and 700 to 850 MHz: (a)–(d) Time-frequency domain energy extreme frequency position of measurement target under the two attenuation lengths of 4.5 and 6; (e), (f) measurement errors for different attenuation length at the two frequency ranges.
基于小波变换的时频融合测距方法的整体测距结果如图6(a)—(f)所示. 可以看出, 在输出功率为2.3 W以及采用APD探测器探测, 调制频率区间为50—200 MHz时, 结合锁相放大器和示波器的最小可分辨强度与相位, 可知其最远探测距离为8.2个衰减长度. 实际探测距离使用均方根(RMS)运算, 可得该测距方法在8个衰减长度内测距精度小于1 cm. 故可知8个衰减长度内的实际测量误差小于0.1个衰减长度. 图 6 基于小波变换的时频融合测距结果 (a)—(f)对于衰减长度为1—8.2的水体, 该测距方法测量的测量误差以及均方根 Figure6. Time-frequency fusion ranging results based on wavelet transform: (a)–(f) Ranging error and root mean square detected by this ranging method for water bodies with attenuation lengths of 1 to 8.2.