A Deep Learning-Based Person Detection Counting Method

Abstract

The invention discloses a method for detecting and counting people based on deep learning, comprising: 1. Periodically collecting video surveillance images, and detecting people based on the FiarMot algorithm; 2. Extracting the area of ​​interest of the detected personnel and calculating the intersection of the area and the monitoring area. And compare; 3. Build an adaptive scale LSTM neural network, fuse the fully connected layer and wavelet transform features and use the SCN classifier for classification, and automatically adjust the network scale according to the entropy loss value evaluation of the detection results. The invention can automatically adjust the network scale according to the evaluation of the entropy loss value of the detection result, so as to realize the self-optimizing adjustment and reconstruction of the video personnel detection model, thereby improving the detection rate of video personnel in designated areas under different environments, and meeting the requirements of accuracy and rapidity. Actual demand.

Description

technical field [0001] The invention belongs to the fields of computer vision, image recognition technology, and deep learning technology, in particular to a method for detecting and counting people based on deep learning. Background technique [0002] Personnel detection and counting has guiding significance for management in some specific occasions. For example, in shopping malls, the estimation of consumers’ personnel can be analyzed, and corresponding consumption strategies can be formulated. Personnel detection and counting in scenic spots can control the number of tourists and avoid crowded and dangerous areas. , improve safety and tourist satisfaction. [0003] However, there are various environmental information, and the current personnel detection and counting model cannot adapt to various complex scenarios, such as coal mine scenarios. Due to the safety regulations of coal mines, personnel at the upper and lower wellheads of the auxiliary shaft can only enter the c...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the shortcomings of the above-mentioned prior art, the present invention proposes a method of personnel detection and counting based on deep learning, in order to solve the problem of high misrecognition rate of personnel detection and counting in complex scenes, and can be based on the entropy loss of detection results Value evaluation automatically adjusts the network scale, realizes automatic adjustment of the detection model and counts personnel detection, thereby improving the detection rate and counting accuracy of personnel in complex backgrounds, and meeting the actual needs of accuracy and speed

Images