Video labeling method and system based on Kalman filtering

Abstract

The invention discloses a video labeling method and system based on Kalman filtering. The method comprises the following steps: carrying out initialization setting on parameters of a Kalman filter; manually determining an initial frame of the position where the object to be labeled is located, and updating the Kalman filter by using the coordinates of the initial frame; predicting a prediction frame of the position of the object to be labeled in the next frame by using a Kalman filter; searching all candidate boxes with the to-be-labeled object in the picture of the current frame by using a selective search algorithm; and matching the prediction frame of the position where the object to be labeled is located with all the candidate frames with the object to be labeled, determining the most matched candidate frame, updating the prediction frame of the object to be labeled in the current frame by using the most matched candidate frame, and labeling according to the updated prediction frame. Compared with a traditional labeling mode, the video labeling efficiency is greatly improved, the efficiency can be improved by 30%-80%, and the labeling speed is effectively increased.

Description

technical field [0001] The present invention relates to the technical field of video tagging, and more specifically, to a video tagging method and system based on Kalman filtering. Background technique [0002] In deep learning, data plays a key role in the prediction results of the neural network model. Data is the "knowledge" learned by the neural network. Usually, a large amount of data is needed to train the neural network to see obvious results. Data labeling is a waste of time. Time-consuming and labor-intensive things, so improving the efficiency of labeling is a key point. [0003] Currently, commonly used data labeling software in the field of target detection include labelimg, cvat, label wizard, etc. Among them, labelimg and label wizard only have basic data labeling functions, cvat is a relatively comprehensive labeling system that supports video labeling, but there are still many shortcomings, such as not supporting prediction of common objects, slow labeling o...

AI Technical Summary

Problems solved by technology

[0005] The present invention proposes a video labeling method and system based on Kalman filtering to solve the problem of how to quickly and accurately perform video labeling

Images