Video target detection method based on deep learning

Abstract

The invention discloses a video target detection method based on deep learning, and is applied to the field of video target detection. According to the method, the convolutional neural network is usedfor extracting image features, and time-is provided; And the spatial feature extraction network is used for extracting spatial context and time context information of the video, fusing image featureswith time and spatial context information, updating a feature map output by the backbone network, and finally inputting the obtained feature map into the detection network to obtain a final detectionresult, thereby giving consideration to the accuracy and real-time performance of target detection. According to the method, the detection accuracy and real-time performance are effectively improved.

Description

technical field [0001] The present invention relates to the field of object detection, and more specifically, to a video object detection method based on deep learning. Background technique [0002] In recent years, deep learning has made unprecedented breakthroughs in the field of computer vision. Through the structure of multi-layer neural network, the overall information of the image is integrated, and the image features are expressed from a higher and more abstract level. At present, the deep learning model based on convolutional neural network (CNN) is widely used in target detection, and has been proved to be better than traditional manual feature methods. [0003] At present, target detection methods are mainly divided into two categories: one is the target detection method based on manual feature extraction, and the other is the target detection method based on deep learning feature extraction. Typical manual features include shape, contour information, etc., and ca...

AI Technical Summary

Problems solved by technology

However, the existing technology has the following defects: (1) The feature extraction method based on manual design usually requires relevant domain knowledge or a large amount of statistical data, thus requiring a huge research cost; affect its accuracy

(2) The calculation amount of the feature extraction method based on deep learning is generally huge, which hinders the application in the actual scene

(3) The current target detection research pays more attention to the detection of static images. Only the redundant information of the video is used to post-process the detection results, or the optical flow is used to recalculate the features, and the time and space context of the video are not fully integrated. information, so video target detection with both accuracy and real-time performance is still an important challenge in current related research and applications.

Images