Video monitoring image human shape identification method

Abstract

The invention discloses a video monitoring image human shape identification method, which comprises the following steps of S1, carrying out Gaussian filtering, noise reduction and image smoothing; s2, processing the video sequence by using a three-frame difference method; s3, expanding the detection processing space range through an expansion method; s4, carrying out contour extraction by adopting a Freeman chain code mode; s5, calculating the curvature value of each point on the contour; s6, acquiring a curvature distribution curve and a curvature space distribution matrix; s7, constructing a learning model; s8, carrying out target threshold matching, and judging the human shape contour. According to the invention, Gaussian filtering and inter-frame difference are carried out on a target; the target post-processing range is expanded by adopting an expansion method, the representation quality of the target can be effectively improved, whether the target is a human shape target or not is judged by adopting a training learning mode in combination with a Learning model, and the defects that the recognition time is long and the recognition time is short in the traditional high-precision method are overcome.

Description

technical field [0001] The invention relates to the technical field of computer algorithms, in particular to a human figure recognition method in video surveillance images. Background technique [0002] At present, there are many methods for humanoid target recognition on surveillance video, which are mainly classified into three categories. One category has unique advantages in feature description, such as using a combination of HOG features and LBP features to describe and recognize humanoid targets to improve the description of humanoid targets. ability, the second category mainly has its advantages in classifiers, such as using classic classifiers such as adaboost and SVM to classify and recognize humanoid targets, and the third category is mainly to build neural network models to identify humanoid targets. The stickiness is better, but optimization can only rely on continuous data training iterations. [0003] There are certain problems in the existing humanoid target ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is that there are certain problems in the existing humanoid target recognition methods, the recognition time is relatively long for those with high precision, and the accuracy for short recognition time is relatively low; and there is no standardized humanoid recognition algorithm at present. There are many related algorithm steps, and a method for human figure recognition in video surveillance images is provided to solve the above problems

Images