摘要
车位检测与车位状态识别是停车场监控与管理的核心技术,而停车场存在复杂的光照背景与差异的环境条件,这对车位的高精度检测与识别提出了挑战。由此提出基于深度学习理论的复杂光照背景车位智能检测与识别方法,经基于线结构光视觉测量的复杂背景车位智能检测方法,获取监控视频图像里复杂光照背景下车位的坐标;由基于深度学习的复杂背景车位智能识别方法里的卷积神经网络提取所定位车位纹理特征后,通过多尺度FPN算法实现纹理特征小目标的训练,完成车位状态智能识别。以某大型露天停车场为例,该方法能够有效检测车位位置、识别空闲车位,且在多种复杂光照背景下,对车位的检测与识别性能均不受干扰,对停车场车位管理存在应用价值。
Parking lot detection and parking spot status recognition are the core technologies of parking lot monitoring and management.The parking lot has a complex lighting background and different environmental conditions,which challenges the high-precision detection and identification of parking spots.This paper proposes an intelligent detection and recognition method for parking Spaces with complex lighting background based on deep learning theory.The intelligent detection method for parking Spaces with complex lighting background based on linear structured light vision measurement is used to obtain the coordinates of parking Spaces with complex lighting background in the surveillance video image.The deep learn-based convolutional neural network in the intelligent recognition method of complex background parking Spaces extracts the positioned parking space texture features.The multi-scale FPN algorithm is used to train the small target of texture features to complete parking spaces’intelligent recognition.Taking a large open parking lot as an example,this method can effectively detect the parking area’s location and identify the vacant parking spaces.Under various complex lighting backgrounds,parking spaces’detection and recognition performance are not disturbed,which has application value for parking space management.
作者
孙桂煌
姜忠海
SUN Guihuang;JIANG Zhonghai(School of Computing and Information Science,Fuzhou Institute of Technology,Fuzhou 350506,China)
出处
《激光杂志》
CAS
北大核心
2021年第4期81-85,共5页
Laser Journal
基金
福建省教育厅科技类科研项目(No.JAT170796)。
关键词
深度学习理论
复杂光照背景
车位智能检测
识别
deep learning theory
complex lighting background
parking lot intelligence detection
recognition
