| According to the statistics of the traffic administration bureau of the Ministry of public security,the number of cars in China is increasing year by year,and the number of urban vehicles is increasing,which makes the problem of parking more and more serious.This paper proposes an automatic parking space detection method to solve people’s difficult parking problems.Based on the widely popularized around view monitor,this paper proposes the design and implementation of a parking space detection algorithm.The main research work includes:(1)Based on ACF(Aggregate Channel Features)to describe parking space corners,multi-scale parking space corner features are established by using a fast feature pyramid,and feature learning is carried out in combination with the idea of Ada Boost(Adaptive Boosting)algorithm,to train a low complexity parking space corner detector that can detect different types,angles,and orientations,which meets translation invariance,rotation invariance,and scale invariance.Then,the speed of the detector is optimized.With a performance loss of 0.2%,the detection speed is increased by nearly 3 times,and the average detection time is 35 ms,which meets the requirements of real-time parking space detection.(2)A general and universal parking information extraction algorithm is proposed.By analyzing the corner area of the parking space detected by the above detector,determine the position of the entrance line and separation line of the current detected parking space,and extract the type,orientation,size,and other information of the parking space.The proposed algorithm has achieved good recognition accuracy in different light and ground environments.At the same time,because the display range of the image data in the current open look around data set is not enough to observe the depth information of the vertical parking space,it is impossible to completely verify the feasibility of the algorithm proposed in this paper.Therefore,we use the around view monitor to collect some image data with a wider display range.We verify the effectiveness of the proposed algorithm on public data sets and self-collected data sets.(3)According to the actual application scenario requirements of parking space detection,the onboard computing platform is selected for algorithm transplantation.TDA2EVM5777,a multi-core heterogeneous on-board computing platform specially designed for advanced driver assistance systems developed by Texas Instruments,is selected to transplant the parking space detection algorithm proposed in this paper.Combined with the characteristics of the multi-core heterogeneous mode of the TDA2EVM5777 platform,a detailed porting scheme is proposed,and tests of the transplanted program are carried out to verify the effectiveness of the porting scheme. |
