A traffic text detection method under the guidance mechanism of expert knowledge

Abstract

The invention discloses a traffic text detection method and system under an expert knowledge guidance mechanism, belonging to the field of traffic text recognition, and mainly includes designing a text detection method based on multi-task deep learning and target tracking strategies; constructing a deep recursive neural network and a multi-frame The text recognition method of the image information fusion mechanism; the text semantic understanding method of exploring the semantic association mechanism guided by expert knowledge; the content of each part is successive and interconnected. Explore an efficient and theoretically grounded solution to improve traffic text detection and recognition performance. The research results of the present invention provide a basis for the in-depth development of research and applications related to assisted driving of automobiles, and have certain reference significance for document analysis and recognition, and other application fields of computer vision.

Description

technical field [0001] The invention relates to the field of traffic text recognition, in particular to a traffic text detection method under an expert knowledge guidance mechanism. Background technique [0002] The appearance of automobiles has greatly facilitated human travel, but with the development of society and the increase of population, the number of automobiles has also shown an explosive growth, which has led to a series of problems such as traffic congestion and frequent traffic accidents, which in turn has caused urban construction and people's lives. endangering property safety. From the development of the automobile industry, we can see that despite the development of automobiles over a century, the driving patterns of automobiles have never undergone essential changes. During the driving process, the driver perceives the traffic environment information through his own visual system and controls the driving state of the vehicle, thus forming a closed-loop sys...

AI Technical Summary

Problems solved by technology

At present, for road information (road lane lines, road edges, traffic signs, etc.), road condition information (obstacles in front of the vehicle, vehicles, pedestrians, etc.), and driver status (driver fatigue information, abnormal driving behavior, etc.) Many effective solutions have been proposed for the visual problems in the driving environment, and their results have been widely published in international authoritative conferences and journals. However, traffic text, an important visual content, has received little attention.

Compared with the visual detection tasks in the driving environment such as pedestrian detection, lane detection, traffic sign detection, and fatigue driving detection, the research work on traffic text detection lags behind significantly.

Images