A parking space detection system and method based on hierarchical pairwise similarity pvanet

Abstract

The invention provides a parking space detection system and method based on hierarchical paired similarity PVAnet, which belongs to the technical field of computer vision. The proposed layered paired similarity PVAnet is mainly composed of two networks. The main framework of the first network was PVAnet. In this network, transfer learning is first used to learn the characteristics of the vehicle, and then the characteristics of the vehicle are integrated into PVAnet. The second network is a paired detection network, which is connected behind the first network to learn the similarity relationship between the detected vehicle center and the spatial position of the parking space center, so that the learned network can be used for parking spaces. effectively identify. The method proposed by the invention can accurately find a parking space, thereby effectively alleviating the difficulty of parking in public places.

Description

technical field [0001] The invention relates to computer vision technology, in particular to a parking space detection system and method based on hierarchical pairwise similarity PVAnet. Background technique [0002] In recent years, the continuous improvement of the social and economic level has caused the number of private cars to increase year by year, making it difficult for the limited public parking resources to meet people's needs. Smart transportation is one of the key projects of the "13th Five-Year Plan" layout. Smart parking is an important part of this major project. Existing schemes all complete the precise detection of parking spaces by installing a large number of geomagnetic induction devices. However, above-ground parking lots account for a large proportion of existing parking lots, and the geomagnetic induction scheme has defects such as high cost and difficult implementation for above-ground parking lots. [0003] In recent years, deep learning has achi...

AI Technical Summary

Problems solved by technology

Faster-rcnn can achieve amazing target detection accuracy, but because it is a two-stage target detection framework, it is slower

SSD simplifies the complexity of the network and becomes a one-stage target detection framework, which greatly improves the target detection speed and can realize real-time target detection. However, due to the simplicity of the network, its accuracy is relatively low

Images