A virtual lane line recognition method, device, equipment, and medium

Abstract

The application discloses a virtual lane line recognition method, device, equipment, and medium. The method at least includes: obtaining road sample images of a plurality of predetermined categories; adjusting the training weights corresponding to the virtual lane line categories, and training the machine learning model so as to focus on training for other categories in the first training stage; adjusting the training weights corresponding to other categories , and retrain the machine learning model to focus on the virtual lane line category training in the second training stage; use the machine learning model trained in the second training stage to identify the virtual lane line in the image to be recognized. This application uses virtual lane lines instead of elements such as fences, convex barriers, and roadsides to define the boundary of one side of the emergency lane, which can simplify the categories required for model training. Not only that, but also by adjusting the categories The training weights enable all categories, including the virtual lane line category, to be fully trained, which is conducive to more accurate recognition of lane lines.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular to a virtual lane line recognition method, device, equipment, and medium. Background technique [0002] In the process of making high-precision maps, lane line recognition is an important content. Common lane lines include white or yellow lines, dashed or solid lines, etc. that appear on the road. The main process of the existing lane line detection scheme includes: binarizing the image to be recognized, and then detecting the straight line by using methods such as Hough transform to identify the lane line. [0003] In practical applications, road boundaries are not all defined by lane lines, and a typical scenario is the emergency lane. It can usually be seen that the right boundary of the emergency lane is often defined by elements such as fences, convex barriers, and roadsides, rather than the lane lines drawn by the traffic control department on the road...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a virtual lane line recognition method, device, equipment, and medium to solve the following technical problems in the prior art: using the existing lane line detection scheme, only conventional lane line detection is performed, and it is difficult to define If the right boundary of the emergency lane is defined by elements such as fences, roadsides, and convex barriers, there are also many elements that are not fully considered, and the elements themselves are difficult to define

Images