Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Parking space identification system based on multi-sensor feature level fusion and identification method thereof
The invention relates to the field of intelligent traffic, and particularly discloses a parking space identification system based on multi-sensor feature level fusion and an identification method thereof. Feature extraction is performed on echo signals of ultrasonic signals of parking spaces through a first convolutional neural network with a cavity convolution kernel; the first convolutional neural network is utilized to enlarge a receptive field on the basis of not losing resolution, so that multi-scale information under high resolution is obtained, the second convolutional neural network is utilized to extract local high-dimensional correlation feature distribution of an outside-vehicle image, and information measurement among internal element sub-dimensions in a high-dimensional space is further carried out, so that multi-scale information of the outside-vehicle image is obtained. The hyperconvex consistency derivative representation of the feature manifolds is performed to perform an appropriate weighted summation between the resulting feature maps based on the derivative consistency of the feature manifolds to obtain an appropriate fused feature map. In this way, parking of people can be more intelligent, and then accidents are avoided.
本申请涉及智能交通的领域，其具体地公开了一种基于多传感器特征级融合的车位识别系统及其识别方法，通过具有空洞卷积核的第一卷积神经网络对车位的超声波信号的回波信号进行特征提取，以在不丢失分辨率的基础上扩大感受野，从而获得高分辨率下的多尺度信息，并且还利用第二卷积神经网络提取出车外图像的局部高维关联特征分布，进一步在高维空间内的内部元素子维度间的信息度量，来进行特征流形的超凸一致性衍生表示，以基于所述特征流型的衍生一致性来进行得到的所述特征图之间的适当的加权求和，以获得适当融合的特征图。这样，就能够使得人们的停车更加智能化，进而避免事故的发生。
Parking space identification system based on multi-sensor feature level fusion and identification method thereof
The invention relates to the field of intelligent traffic, and particularly discloses a parking space identification system based on multi-sensor feature level fusion and an identification method thereof. Feature extraction is performed on echo signals of ultrasonic signals of parking spaces through a first convolutional neural network with a cavity convolution kernel; the first convolutional neural network is utilized to enlarge a receptive field on the basis of not losing resolution, so that multi-scale information under high resolution is obtained, the second convolutional neural network is utilized to extract local high-dimensional correlation feature distribution of an outside-vehicle image, and information measurement among internal element sub-dimensions in a high-dimensional space is further carried out, so that multi-scale information of the outside-vehicle image is obtained. The hyperconvex consistency derivative representation of the feature manifolds is performed to perform an appropriate weighted summation between the resulting feature maps based on the derivative consistency of the feature manifolds to obtain an appropriate fused feature map. In this way, parking of people can be more intelligent, and then accidents are avoided.
本申请涉及智能交通的领域，其具体地公开了一种基于多传感器特征级融合的车位识别系统及其识别方法，通过具有空洞卷积核的第一卷积神经网络对车位的超声波信号的回波信号进行特征提取，以在不丢失分辨率的基础上扩大感受野，从而获得高分辨率下的多尺度信息，并且还利用第二卷积神经网络提取出车外图像的局部高维关联特征分布，进一步在高维空间内的内部元素子维度间的信息度量，来进行特征流形的超凸一致性衍生表示，以基于所述特征流型的衍生一致性来进行得到的所述特征图之间的适当的加权求和，以获得适当融合的特征图。这样，就能够使得人们的停车更加智能化，进而避免事故的发生。
Parking space identification system based on multi-sensor feature level fusion and identification method thereof
基于多传感器特征级融合的车位识别系统及其识别方法
HAN ZHONGJUN (Autor:in) / ZHAO XIAOLONG (Autor:in)
22.11.2022
Patent
Elektronische Ressource
Chinesisch
Parking space identification unlocking system and control method thereof
| Europäisches Patentamt | 2022
Parking space informatization system based on multi-source heterogeneous sensor fusion
| Europäisches Patentamt | 2023
Charging pile parking space automatic identification parking lock system
| Europäisches Patentamt | 2022
Parking space line identification method based on vision
| Europäisches Patentamt | 2021
Multi-parking space intelligent robot parking equipment and parking method thereof
Europäisches Patentamt | 2020