A method for moving target detection and smooth following based on video images

Abstract

The invention provides a method for detecting and smoothly following moving targets based on video images, which includes the following steps: preprocessing the image sequence to obtain a foreground image; using the contour recognition method to filter out all moving targets; and quickly reducing the dimensionality Calculate the horizontal position of the moving object; smooth the motion state of the target through the boundary layer strategy; generate the angular velocity curve of the model robot following the movement. The advantage of the present invention is that the algorithm is simple and easy to implement, can quickly identify moving targets, and has low requirements on the hardware configuration environment. It can be widely used in mannequin robot vision systems, while realizing the mannequin robot smoothly following the movement of the moving target while also avoiding The phenomenon of jitter is eliminated.

Description

technical field [0001] The invention relates to the technical field of video image processing, in particular to a method for detecting and smoothly following a moving target based on video images. Background technique [0002] Traditional window models can only statically display clothes, and using model robots for fashion displays can attract customers to stop by interacting with customers, thereby greatly increasing passenger flow. The invention provides a method for detecting and smoothly following a moving target based on a video image, which is applied to the vision system of a model robot to realize its recognition and smooth following of a moving target. When a pedestrian appears or passes in front of the model, the model Robots can perform turn-following or other forms of interaction. [0003] The more common video-based target tracking method is to detect and track the determined target through feature extraction and target tracking. However, the pedestrians faced ...

AI Technical Summary

Problems solved by technology

[0003] The more common video-based target tracking method is to detect and track the determined target through feature extraction and target tracking. However, the pedestrians facing the models in the shop window are constantly changing, and the determined target cannot be preset in advance.

The more common methods such as neural network, deep learning and SVM based on machine learning train the classifier through a large number of pedestrian or face samples, and then detect the video image, but have higher requirements for the computing speed of the hardware environment

Images