德清县五水共治遥感监测识别数据
收藏资源简介：
用于实现对德清县五水共治遥感监测中问题点位的高效精准识别，包括水面蓝藻、垃圾等水面环境物问题。自动识别问题点位的坐标，便于确定需要进行实地处理的位置。自动识别点位问题类型，精准发现水面环境物。为后续的人员管理与问题派发提供方便，有利于网格员及时掌握自己管理区域的情况。解决了网格员在传统人工巡查中难以发现问题，容易忽略问题，巡查范围太大，隐蔽地段、偏远地区与恶劣环境中不易巡查等问题。将系统识别出的问题点位派发给网格员，使网格员有依据、有目标的实地确认，极大的提高了发现问题的效率，节约人员时间与人工成本，避免网格员出现无效的巡查。基于无人机航拍采集的五水共治遥感影像数据，通过YOLO算法进行实时目标检测。首先将单元神经网络应用于2024年9月的遥感影像，将图像分割成19x19的单元格，每个单元神经网络负责预测K个单元格。预测每个区域的概率，所有单元格上具有最大概率的类被选择并分配给特定的网格单元，生成预测点坐标（x，y），坐标系为CGCS2000，坐标为东经、北纬。 在预测类概率后，进行NMS运算，来消除不必要的锚点。算法识别下一个最高类别概率的边界框，并进行相同的运算过程，直到剩下所有不同的边界框。算法输出所需的要素，并显示各个类的边界框的细节。 抽取部分样本进行识别准确度验证，小于0.6视为识别错误，显示为FALSE；一般样本的识别准确度在0.8至1之间，大于等于0.6视为识别正确，显示为TRUE。通过判断结果正确或错误来纳入或排除数据，将识别正确的点位进行五水共治问题类型划分，仅纳入类别为水面环境物的问题。最后将纳入的点位坐标、问题类型等信息自动上传至五水共治智能监管平台，获得德清县五水共治遥感监测识别数据。
This dataset is designed for efficient and accurate identification of problem spots in remote sensing monitoring of Five-Water Governance in Deqing County, covering issues of surface environmental objects such as surface cyanobacteria and garbage. It automatically identifies the coordinates of problem spots to facilitate the confirmation of locations requiring on-site treatment, and recognizes the problem types of the spots to accurately detect surface environmental objects. This facilitates subsequent personnel management and task assignment, enabling grid workers to timely grasp the status of their respective management areas. It addresses the challenges faced by grid workers in traditional manual patrols, including difficulty in detecting problems, proneness to overlooking issues, overly large patrol scope, and difficulties in patrolling hidden areas, remote regions and harsh environments. By assigning the problem spots identified by the system to grid workers, it allows them to conduct on-site verification with clear basis and targets, greatly improving the efficiency of problem detection, saving personnel time and labor costs, and eliminating invalid patrols for grid workers. The workflow is based on remote sensing image data of Five-Water Governance collected via UAV aerial photography, with real-time object detection implemented using the YOLO algorithm. First, the unit neural network is applied to the remote sensing images from September 2024, which are split into 19×19 grid cells. Each unit neural network is responsible for predicting K bounding boxes per assigned cell. The model predicts the category probability for each region, then selects the class with the maximum probability across all grid cells and assigns it to a specific grid cell, generating the predicted spot coordinates (x, y) in the CGCS2000 coordinate system, where the coordinates are expressed as east longitude and north latitude. After predicting the category probabilities, a Non-Maximum Suppression (NMS) operation is performed to eliminate redundant anchor boxes. The algorithm identifies the bounding box with the next highest category probability and repeats the same process until all distinct bounding boxes remain. Finally, the algorithm outputs the required elements and displays the detailed information of the bounding boxes for each category. A portion of samples are selected for recognition accuracy verification. Samples with an accuracy score lower than 0.6 are classified as recognition errors and marked as FALSE; samples with an accuracy score greater than or equal to 0.6, including general samples with an accuracy score ranging from 0.8 to 1, are regarded as correctly recognized and marked as TRUE. Data is included or excluded based on the correctness of the recognition result. The correctly recognized spots are classified according to the problem types of Five-Water Governance, and only those categorized as surface environmental objects are included. Finally, the information of the included spots such as coordinates and problem types is automatically uploaded to the Five-Water Governance Intelligent Supervision Platform, thereby generating the remote sensing monitoring and recognition data for Five-Water Governance in Deqing County.
