An automatic parking space detection method based on deep learning

Abstract

The invention relates to the technical field of automatic driving, in particular to an automatic parking space detection method based on deep learning. The method comprises the steps of summing the loss function obtained on the basis of deep learning and the cost function of the template obtained on the basis of template matching to obtain a total cost function, and searching the total cost function by taking the position of the point with the maximum probability of any model as an initial value to obtain the position of the parking space. Compared with the prior art, the method has the advantages that the method is easy to implement; according to the method, template matching and deep learning are combined to carry out parking space detection in the image; the method is effective under the conditions of good illumination conditions and good splicing effects, and can realize robust and accurate parking space detection under the conditions of complex light sources, road surface reflection, interference in parking spaces, inaccurate splicing top views and the like. And the calculation complexity of distance transformation and Canny edge detection is very low, the calculation amount is very small, and the method can be operated on an embedded platform in real time, so that the method has a direct application value.

Description

technical field [0001] The invention relates to the technical field of automatic driving, in particular to an automatic parking space detection method based on deep learning. Background technique [0002] NHTSA divides unmanned driving technology into 6 levels, which are 0-5. Among them, L0 is a general vehicle with full manual control, L1 level is also called assisted driving, which can realize simple acceleration and deceleration, and L2 level is also called partial automatic driving. It can realize all the content of L1 level and also realize automatic parking, while L4 and L5 levels can realize fully automatic driving, the difference is that L4 level can only realize fully automatic driving under specific roads and weather, while L5 level can adapt All terrain and all weather. Vision-based automatic parking is an important application of unmanned driving. Since the vision is easily disturbed during the reversing process, such as light interference, sewer fences, etc., t...

AI Technical Summary

Problems solved by technology

Vision-based automatic parking is an important application of unmanned driving. Since the vision is easily disturbed during the reversing process, such as light interference, sewer fences, etc., the vision-based parking space detection during the reversing process is not robust enough.

In the prior art, for example, my country's Patent Publication No.: CN109086708A discloses a method for detecting parking spaces based on deep learning. However, using deep learning alone to detect parking spaces is not accurate enough in many scenarios. , and it is easy to produce wrong detection results when the light source is complex and the road surface is reflective.

However, my country's Patent Publication No. CN105160322A discloses a method for identifying vacant parking spaces in an outdoor parking lot based on aerial images. The identification of parking spaces is realized through template matching, but the robustness is still not high and the amount of calculation is relatively large.

Images