生物远程感知和驱赶技术在运检工作中的应用研究

Research on Application of Biometric Remote Sensing and Driving Technology in Operation and Inspection

摘要: 本文针对变电站日常维护过程中生物体异常入侵检测和驱逐的普遍需求，尤其是针对小型生物体难以通过单一检测手段准确识别的问题，提出了通过可见光和热红外相机组合的方式对小动物入侵进行识别的方法。该方法通过可见光相机画面的CNN智能识别，得到多组疑似小动物的识别框，然后映射到热红外相机的画面中，依次对每个识别框进行温度阈值检测并判断是否真的存在生命体。如果确实识别到小动物，那么通过物联网端口自动调用超声波发射设备进行驱赶操作。该方法解决了小型生物体视觉识别的误报率高，稳定性差的常见问题，并且可以与其它传感器识别和驱赶方案自然结合，适合应用在各种典型的电力运维场景当中。

Abstract: In this paper, we study a scheme to identify small animal intrusion by a combination of visible and thermal infrared cameras, which solves the common need for abnormal organism intrusion detec-tion and expulsion in the daily maintenance process of substations, especially for small organisms that are difficult to be accurately identified. The visible camera is used to identify multiple groups of suspected small animal frames using CNN method. Those groups are then mapped to the thermal infrared camera, and subjected to a temperature threshold detection task to determine whether there is a real animal. If identified, an ultrasonic transmitter is automatically invoked via the IoT port to carry out the repelling operation. The solution is of high stability, high security and low false alarm rates of visual recognition of small organisms, making it suitable for application in a variety of typical power operation and maintenance scenarios.