无人机AI智能识别环卫工人数据

针对城市道路清洁作业，无人机凭借 70 至 100 米的中低空优势，结合高分辨率传感器与 AI 算法，能快速识别作业中的环卫工人。给环卫企业动态化管理环卫工人提供数据支撑。可广泛应用于：
1、环卫工人作业范围与时长统计
针对环卫工人分散作业的特点，无人机通过人形识别与特征提取，精准统计各路段环卫工人的在岗数量、作业时长指导管理人员动态调配人力。识别数据同步至管理中心，还能统计各区域作业密度，优化人员排班，提升整体作业效率。
2、环卫工人安全防护监察
通过 AI 算法识别环卫工人是否规范穿戴反光服、安全帽等防护装备，对未按规定着装者实时推送预警至管理中心并发送警告声至环卫工人佩戴设备，联动地面督导人员及时纠正，降低作业风险。
3、环卫工人作业轨迹监督
无人机可识别环卫工人是否按预定轨迹作业。是否在严禁作业时段或危险区域作业，发现违规立即警告，同步标记位置信息，记入日志。1、数据来源: 数据来源于本企业无人机智能巡查系统。2、高分辨率图像通过无人机采集，记录丰富元信息，包括图像标识ID、图像分辨率(pix)、相机型号、记录时间、文件路径、焦距(mm)、经纬度、海拔(m)、边界框组、置信度阈值、置信度、一级标签和二级标签，经人工清洗剔除噪声，确保数据可靠性。数据采用两级标签：初始细化目标类型标为二级标签（环卫工人），记录边界框及属性，后映射为一级标签（人员）。3、算法基于YOLOv8m，集成SAHI通过图像分片优化小目标检测，结合迁移学习加载预训练权重。预训练权重使用在COCO数据集上预训练的模型参数yolov8m.pt，适用于通用目标检测任务。微调时，保留骨干网络低层次特征提取层权重，冻结部分层以防过拟合，调整检测头参数，设置学习率0.01、批量大小8。4、在推理阶段，设定置信度阈值为0.45，目标置信度由模型输出，反映目标检测的可信水平。仅保留高于此阈值的检测结果作为目标，即置信度输出大于等于0.45的检测框视为正样本（环卫工人），小于0.45的检测框视为背景负样本。

For urban road cleaning operations, UAVs can rapidly identify sanitation workers during their work shifts by leveraging their medium and low-altitude operational advantage (70 to 100 meters), combined with high-resolution sensors and AI algorithms, providing data support for sanitation enterprises to dynamically manage their workers. It can be widely applied in the following scenarios:
1. Statistics on the operation scope and duration of sanitation workers
Given the decentralized operation characteristics of sanitation workers, UAVs can accurately count the number of on-duty workers and their working duration on each road section via human shape recognition and feature extraction, to guide managers in dynamically allocating labor resources. The recognized data is synchronized to the management center, and the operation density of each area can also be counted to optimize staff scheduling and improve overall operation efficiency.
2. Safety protection supervision for sanitation workers
AI algorithms are used to identify whether sanitation workers wear protective equipment such as reflective vests and safety helmets in accordance with regulations. For those who fail to dress as required, early warnings will be pushed to the management center in real time, and warning sounds will be sent to the devices worn by sanitation workers. Ground supervisors will be linked to correct the situation in time, thereby reducing operation risks.
3. Operation trajectory supervision for sanitation workers
UAVs can identify whether sanitation workers operate according to the predetermined trajectory, whether they operate during forbidden operation periods or in dangerous areas. If violations are found, immediate warnings will be issued, and the location information will be marked synchronously and recorded in the log.

### Data Details
1. Data Source: The data comes from the UAV intelligent inspection system of our company.
2. High-resolution images are collected by UAVs, with rich metadata recorded, including image ID, image resolution (pix), camera model, recording time, file path, focal length (mm), longitude and latitude, altitude (m), bounding box group, confidence threshold, confidence, first-level labels and second-level labels. Noise is eliminated through manual cleaning to ensure data reliability. A two-level labeling scheme is adopted: the initially refined target type is marked as the second-level label (sanitation worker), with bounding boxes and attributes recorded, and then mapped to the first-level label (person).
3. The algorithm is based on YOLOv8m, and integrates SAHI to optimize small-object detection via image slicing. Pre-trained weights are loaded via transfer learning. The pre-trained weights use the model parameters yolov8m.pt pre-trained on the COCO dataset, which is suitable for general object detection tasks. During fine-tuning, the weights of the low-level feature extraction layers of the backbone network are retained, some layers are frozen to prevent overfitting, the detection head parameters are adjusted, with the learning rate set to 0.01 and batch size set to 8.
4. In the inference phase, the confidence threshold is set to 0.45. The target confidence output by the model reflects the credibility level of object detection. Only detection results above this threshold are retained as valid targets, that is, detection boxes with confidence output greater than or equal to 0.45 are regarded as positive samples (sanitation workers), while those with confidence less than 0.45 are regarded as background negative samples.

搜集汇总

背景与挑战

背景概述

该数据集包含502条无人机采集的环卫工人图像数据，每日更新，记录丰富元信息如位置、时间、置信度等，用于AI智能识别。其应用支持环卫工人作业统计、安全防护和轨迹监督，基于YOLOv8m算法优化检测精度，提升城市环卫管理效率。