Crowd Density Spectrum Estimation Method Based on Local Texture Features

Abstract

The invention provides a method for estimating crowd density spectrum based on local texture features. The SIFT key point detector can be used to effectively detect the area where pedestrians are located in the image, and the SVM classifier is used to classify each key point area, and each key point can be effectively obtained. Crowd density in the point area. In addition, the present invention also provides a texture feature extraction method, which expresses the pedestrian density of the image block by extracting the gray level co-occurrence matrix feature of the local texture pattern, which is simple and effective.

Description

technical field [0001] The invention relates to a video image processing technology, in particular to a technology for real-time detection of crowd density in a large public place. Background technique [0002] Most of the previous crowd density estimation methods extract the global features of the image and estimate the crowd density of the entire image. However, this method has two serious disadvantages: 1. Pedestrians may only appear in a certain part of the image, for example: in a picture containing roads, pedestrians may only appear on the sidewalk; 2. In practical applications, people More concerned is the crowd density in a specific area, for example: in a movie theater, the crowd density at the entrance is more important than the crowd density sitting in the waiting area. Therefore, in order to better adapt to various application scenarios, it is necessary to estimate the crowd density spectrum of the image. [0003] Crowd density spectrum estimation needs to comp...

AI Technical Summary

Problems solved by technology

However, this method has two serious disadvantages: 1. Pedestrians may only appear in a certain part of the image, for example: in a picture containing roads, pedestrians may only appear on the sidewalk; 2. In practical applications, people More concerned about the crowd density in a specific area, for example: in a movie theater, the crowd density at the entrance is more important than the crowd density sitting in the waiting area

This method consumes a lot of memory and time due to the large number of extracted feature blocks and high feature dimensions, and uses a tree structure such as a random forest to store features, making it difficult to meet real-time requirements.

Images