A pedestrian detection method based on an adaptive feature channel

Abstract

The invention discloses a pedestrian detection method based on a self-adaptive feature channel, and the method comprises the following steps of 1, obtaining a Calch pedestrian data frame, and capturing a training set and a verification set based on the data frame; 2, zooming images of the training set and the verification set; 3, taking the standardized image as model input data; 4, designing a convolutional neural network based on a Caffe open-source deep learning framework, and outputting whether an image contains information about whether a pedestrian exists or not, position information ofthe pedestrian and the possibility that the pedestrian exists at the position; 5, calculating the loss of the output information of the model and the corresponding annotation information, and optimizing and training the model; and 6, detecting pedestrians existing in the real scene image by using the optimized model parameters to obtain a corresponding pedestrian detection frame. The method is based on the convolutional neural network and adopts the transfer learning and multi-task joint learning technology, the end-to-end pedestrian detection method is achieved, and the important technical support is provided for follow-up operation of pedestrian detection.

Description

technical field [0001] The invention relates to a convolutional neural network and a pedestrian detection technology, in particular to a pedestrian detection method based on an adaptive feature channel of the convolutional neural network. Background technique [0002] Pedestrian detection is one of the important research topics in computer vision, and its research results have an important impact on other visual tasks. Pedestrian detection has been widely used in vehicle assisted driving, intelligent video surveillance and human-computer interaction systems, but in some complex scenarios, there is still a problem of low detection performance. Pedestrians are flexible objects, and their different postures and appearances will cause the gap between pedestrians to become larger to a certain extent, making it difficult to detect pedestrians. [0003] With the steady development of deep learning and machine learning, the use of convolutional neural networks to learn target featu...

AI Technical Summary

Problems solved by technology

Pedestrian detection has been widely used in vehicle assisted driving, intelligent video surveillance and human-computer interaction systems, but in some complex scenes, there is still the problem of low detection performance

Pedestrians are flexible objects, and their different postures and appearances will cause a larger gap between pedestrians to a certain extent, making it difficult to detect pedestrians

Images