An automatic driving scene key target detection and extraction method based on deep learning

Abstract

The invention relates to an automatic driving scene key target detection and extraction method based on deep learning, and is characterized in that the method carries out key target detection, semantic annotation and target frame selection in a driving scene based on an improved single multi-frame detector, and the training process of the improved single multi-frame detector specifically comprisesthe steps: 1) obtaining a training set; 2) generating a prior frame, and matching the prior frame with a real target with class labels and frame labels of corresponding pictures to form positive andnegative samples; And 3) iteratively updating the model parameters of the single multi-frame detector by utilizing the loss function, wherein the iterative updating is realized based on momentum optimization gradient updating. Compared with the prior art, the method has the advantages of high detection precision, capability of detecting a small target and the like.

Description

technical field [0001] The invention relates to a method for detecting an object in an automatic driving scene, in particular to a method for detecting and extracting key objects in an automatic driving scene based on deep learning. Background technique [0002] With the rapid development of deep learning technology in the field of image and video, the advantages of a series of target detection algorithms related to deep learning are gradually emerging. When dealing with driving vision scenes or other more complex detection scenes, target recognition and detection algorithms based on deep learning and vision have considerable application space. [0003] Some target detection algorithms based on deep network have been proposed in the visual perception technology of unmanned driving. Applying deep learning to target detection in driving scenarios has become a current research hotspot. Early studies applied multi-scale convolutional neural networks to traffic sign classificati...

AI Technical Summary

Problems solved by technology

[0004] First, most of the existing research focuses on the detection of single-type targets such as pedestrians, traffic signs or obstacles, and lacks a general key target recognition framework.

[0005] Second, part of the application of visual deep learning to unmanned visual recognition cannot achieve accurate automatic target selection, and the detection effect for long-distance small targets is not ideal

Images