住宅电梯内电动车进入阻止预警数据

应用于防止电动车进入住宅电梯的预警监测，通过建立标准数据库和监测进入电梯电动车特征数据，通过标准信息对比识别违法进入电梯的电动车实体，预警提示和阻止电动车通过电梯进入居民住宅或楼道，确保有效阻止电动车通过电梯进入居民楼进行充电，有效防止电动车在居民家中或楼道充电意外火灾而导致发生居民楼重特大火灾事故隐患的发生。住宅电梯内电动车进入阻止预警算法描述：一、电动车数据建模。基于灰度图物体识别（SIFT/SURF），建立电动车图像金字塔，形成三维的图像空间，通过Hessian矩阵获取每层局部极大值，在极值点周围26个点进行NMS得到粗略的特征点；使用二次插值法得到精确特征点所在的层。二、数据模型加工。在特征点选取一个与尺度相应的邻域，求出主方向，其中SIFT采用一个正方形邻域内统计所有点的梯度方向，找到80%以上的方向作为主方向；SURF选择圆形邻域，使用活动扇形方法求出特征点主方向，以主方向对齐即完成旋转不变。三、数据模型比对。电动车进入电梯，预警摄像头对电动车进行符合像素要求的拍照，以主方向为轴在每个特征点建立坐标，SIFT在特征点选择一块大小与尺度相应的方形区域，分成16块，统计每一块沿着8个方向占比，特征点形成128维特征向量，对进入电梯的电动车图像归一化则完成强度不变；SURF分成64块，统计每一块的dx，dy，|dx|，|dy|的累积和，形成128维向量，进行归一化完成对比度不变与强度不变，完成电梯电动车的识别并发出预警信息；预警信息同步传入电梯控制系统，电梯停止运行。

Applied to early warning and monitoring for preventing electric vehicles from entering residential elevators: A standard database and feature data of electric vehicles entering elevators are established, and illegally entering electric vehicle entities are identified via comparison with standard information. Early warnings are issued and electric vehicles are blocked from entering residential buildings or corridors via elevators, effectively preventing electric vehicles from being charged in residential spaces or corridors, thus eliminating the hidden danger of major fire accidents caused by accidental fires from electric vehicle charging in residents' homes or corridors.

The algorithm description for early warning and blocking of electric vehicles in residential elevators is as follows:
1. Electric Vehicle Data Modeling
Based on grayscale image object recognition (SIFT/SURF), an image pyramid of electric vehicles is constructed to form a 3D image space. Local maxima of each layer are extracted via the Hessian matrix, and rough feature points are obtained by performing Non-Maximum Suppression (NMS) on 26 neighboring points around the extreme points. Quadratic interpolation is used to determine the layer where the precise feature points are located.
2. Data Model Processing
A scale-adaptive neighborhood is selected at each feature point to calculate the principal direction. For SIFT, gradient directions of all points within a square neighborhood are counted, and directions accounting for over 80% of the total are selected as the principal direction. For SURF, a circular neighborhood is adopted, and the principal direction of the feature point is calculated using the moving sector method, with rotation invariance achieved by aligning with the principal direction.
3. Data Model Matching
When an electric vehicle enters the elevator, the early warning camera captures images of the electric vehicle that meet the pixel requirements. Coordinates are established for each feature point with the principal direction as the axis. For SIFT, a square region matching the scale is selected at the feature point, divided into 16 sub-regions, and the proportion of each sub-region along 8 directions is counted, forming a 128-dimensional feature vector for the feature point. Normalization of the captured electric vehicle image achieves intensity invariance. For SURF, the region is divided into 64 sub-regions, and the cumulative sums of dx, dy, |dx|, and |dy| of each sub-region are counted to generate a 128-dimensional vector. Normalization is performed to achieve both contrast invariance and intensity invariance, completing the recognition of electric vehicles in the elevator and triggering early warning information. The early warning information is synchronously transmitted to the elevator control system, and the elevator stops operating.

搜集汇总

数据集介绍

特点

该数据集包含251条住宅电梯内电动车进入的预警数据，每日更新，用于通过算法识别和阻止电动车进入电梯，以减少火灾隐患。数据来源于企业，并通过浙江省知识产权区块链公共存证平台存证。