Static target segmentation method and system

Abstract

The invention discloses a static target segmentation method and system, and the method comprises the steps: 1, obtaining a current frame image and a reference frame image, inputting the current frame image and the reference frame image into a texture extraction network module, and obtaining the texture features of the corresponding images; step 2, using an attention calculation module to cut the texture features of the current frame and the reference frame into blocks, then calculating the similarity of each block, and obtaining soft and hard attention based on the calculation of the similarity; 3, inputting the current frame image and the reference frame image into a coding network module to obtain corresponding coding features, obtaining attention features through an attention feature generation module by using the reference frame coding features and hard attention, and obtaining final attention features according to the attention features and the current frame coding features; and 4, constructing a decoding network module, and obtaining a final static target segmentation result of the current frame image based on the current frame coding feature and the final attention feature. According to the method, the segmentation precision is ensured, and meanwhile, the overall logic processing speed is increased.

Description

Technical field [0001] The present invention belongs to the field of static target segmentation, involving a static target segmentation method and system. Background technique [0002] Based on visual static target segmentation, such as car lines, sidewalks, traffic signs, environmental obstacles, etc., are important technologies in unmanned visual perceptions, with lane identification as an example, current research results in single frame data Using, the stability of the detection result is difficult to ensure that complex post-processing logic is required to increase the difficulty of developing deployments, while increasing the difficulty of development deployment. [0003] Single frame car line identification technology is mainly divided into segmentation-based, based on anchor-based, the four categories of parameter prediction, parameter prediction, parameter prediction; It can produce more precise and rich identification results, but the power is higher, so many researcher...

AI Technical Summary

Problems solved by technology

[0002] Vision-based static target segmentation, such as lane lines, sidewalks, traffic signs, environmental obstacles, etc., is an important technology in the visual perception of unmanned driving. Taking lane line recognition as an example, the current research results in this field focus on single-frame data. Therefore, it is difficult to guarantee the stability of the detection results from the model level, and complex post-processing logic is required, which increases the difficulty of development and deployment, and also brings additional computing overhead

Images