Li Bingchen, Mo Di, Wang Peisi, Gan Nan, Lin Miao, Wang Ran, Li Shiqiang
Appl Opt. 2021 Sep 20;60(27):8328-8335. doi: 10.1364/AO.431516.
Frequency-modulated continuous-wave lidar realizes 4D (three-dimensional space and velocity) imaging of the scene by emitting positive and negative frequency sweep laser signals. The premise of it is to identify the frequency points corresponding to the same target in the positive and negative sweep echo signals. For dechirp receiving, there is usually one peak in the frequency spectrum of the positive and negative sweep signals, respectively. Therefore, it is easy to identify and match the peaks. But in a complex environment, the laser beam will irradiate multiple targets at the same time. In addition, beam scanning and target motion cause the echo spectrum to broaden. The above reasons make it extremely difficult to identify and match peaks in practice. To solve this problem, the waveform-matching algorithm based on the skeleton tree is first applied to multitarget echo pairing. The basic idea of the algorithm is to quantify the target echo hierarchically to generate a skeleton tree. The generation of nodes is based on the relative amplitude of waveform peaks and reflects the characteristics of wave crests nesting. Then the similarity of the signal is determined by comparing the distance between the two signal waveform feature trees. Finally, the waveforms are matched in terms of similarity. To further substantiate the role of the proposed algorithm, imaging experiments and related comparative data for different targets have been completed. The results show that the accuracy of matching processed by the algorithm exceeds 90%, which is improved by about 50% compared with not using the algorithm for the target whose overlapping part accounts for a large proportion of itself.
调频连续波激光雷达通过发射正负扫频激光信号实现场景的4D（三维空间和速度）成像。其前提是在正负扫频回波信号中识别对应同一目标的频率点。对于去啁啾接收，正负扫频信号的频谱中通常分别有一个峰值。因此，很容易识别和匹配这些峰值。但在复杂环境中，激光束会同时照射多个目标。此外，光束扫描和目标运动会导致回波频谱展宽。上述原因使得在实际中识别和匹配峰值极其困难。为解决这个问题，基于骨架树的波形匹配算法首次应用于多目标回波配对。该算法的基本思想是对目标回波进行分层量化以生成骨架树。节点的生成基于波形峰值的相对幅度，反映波峰嵌套的特征。然后通过比较两个信号波形特征树之间的距离来确定信号的相似度。最后，根据相似度对波形进行匹配。为进一步证实所提算法的作用，已完成针对不同目标的成像实验及相关对比数据。结果表明，该算法处理后的匹配准确率超过90%，对于重叠部分占自身比例较大的目标，与不使用该算法相比提高了约50%。
