Mosquito-borne diseases are a major challenge for Human health as they affect nearly 700 million people every year and result in over 1 million deaths. Reliable information on the evolution of population and spatial distribution of key insects species is of major importance in the development of eco-epidemiologic models. This paper reports on the remote characterization of flying mosquitoes using a continuous-wave infrared optical remote sensing system. The system is setup in a controlled environment to mimic long-range lidars, mosquitoes are free flying at a distance of ~ 4 m from the collecting optics. The wing beat frequency is retrieved from the backscattered light from mosquitoes transiting through the laser beam. A total of 427 transit signals have been recorded from three mosquito species, males and females. Since the mosquito species and gender are known a priori, we investigate the use of wing beat frequency as the sole predictor variable for two Bayesian classifications: gender alone (two classes) and species/gender (six classes). The gender of each mosquito is retrieved with a 96.5% accuracy while the species/gender of mosquitoes is retrieved with a 62.3% accuracy. Known to be an efficient mean to identify insect family, we discuss the limitations of using wing beat frequency alone to identify insect species.

中文翻译：

---

用于监测飞蚊的光学遥感、性别识别和物种识别讨论

蚊媒疾病是人类健康面临的重大挑战，因为它们每年影响近 7 亿人，并导致超过 100 万人死亡。关于关键昆虫物种的种群进化和空间分布的可靠信息对于生态流行病学模型的开发具有重要意义。本文报告了使用连续波红外光学遥感系统对飞蚊的远程表征。该系统设置在受控环境中以模拟远程激光雷达，蚊子可以在距收集光学器件约 4 m 的距离自由飞行。翅膀拍频是从蚊子穿过激光束的反向散射光中提取的。总共记录了 427 个过境信号，来自三种蚊子，雄性和雌性。由于蚊子的种类和性别是先验已知的，我们研究了使用翅膀拍频作为两个贝叶斯分类的唯一预测变量：单独的性别（两类）和物种/性别（六类）。每只蚊子的性别检索准确率为 96.5%，而蚊子的种类/性别检索准确率为 62.3%。已知是识别昆虫家族的有效手段，我们讨论了单独使用翅膀拍频来识别昆虫物种的局限性。