A multi-scale pedestrian detection method and device

Abstract

Embodiments of the present invention provide a multi-scale-based pedestrian detection method and device, wherein the method includes: inputting an image to be detected into a feature extraction model, and outputting a set of image features; the feature extraction model is based on a Resnet network and includes multiple layers The neural network of the convolution units in series, correspondingly, the image feature set is composed of the image features output by each convolution unit in the multi-layer series of convolution units; the image feature set is input to the feature fusion model, and the fusion feature is output; The fusion feature is input to the feature regression classification model, the regression result and classification result of the fusion feature are output, and the pedestrian detection result is obtained based on the regression result and classification result. The method and device provided by the embodiments of the present invention, while realizing multi-scale pedestrian detection, simplifies the feature extraction model, increases the flexibility of pedestrian detection, reduces computing overhead, avoids excessive resource occupation, and improves multi-scale pedestrian detection. Accuracy of pedestrian detection.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of computer vision, and in particular to a multi-scale-based pedestrian detection method and device. Background technique [0002] Pedestrian detection can be defined as: judging whether the input picture (or video frame) contains pedestrians, and if so, giving location information. Pedestrian detection is a research hotspot in the field of computer vision. It is the first step in applications such as vehicle assisted driving, intelligent video surveillance, and human behavior analysis. In recent years, it has also been used in emerging fields such as aerial images and victim rescue. [0003] In real scenes, due to factors such as the height of the monitoring equipment, the shooting angle, and the distance between the pedestrian and the monitoring equipment, the size of the pedestrians captured by the monitoring equipment is different, and the size span is large, which brings difficulti...

AI Technical Summary

Problems solved by technology

[0004] Embodiments of the present invention provide a multi-scale pedestrian detection method and device to solve the problems of high computational overhead or complex operation in existing multi-scale pedestrian detection methods

Images