Pedestrian detection method based on multi-scale convolution characteristics

Abstract

The invention discloses a pedestrian detection method based on multi-scale convolution characteristics, and the method comprises the steps: obtaining a to-be-identified image, converting the to-be-identified image into a set size, and storing the set size as a converted image; inputting the converted image into a VGG16 network model for feature extraction, storing the output of the last down-sampling layer as a first feature map, and storing the output of the previous convolutional layer of each down-sampling layer as a second feature map; inputting the first feature map into an area recommendation network to obtain a foreground recommendation area; Intercepting an area, corresponding to the recommended area, of the converted image as a sub-image, and inputting the sub-image into a VGG16 network model to obtain a third feature map of each sub-image; intercepting feature maps of areas, corresponding to the third feature map, on the first feature map and all the second feature maps; splicing all the feature maps intercepted by the same third feature map, and inputting the spliced feature maps into a recognition network for recognition to obtain the probability that the recommended area is the pedestrian.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a pedestrian detection method based on multi-scale convolution features. Background technique [0002] Due to the needs of public area management and security, intelligent video surveillance has become one of the important applications of computer vision. The key step of intelligent video surveillance is target detection, especially pedestrian detection. Accurate target detection provides a good foundation for subsequent intelligent analysis, such as target tracking, target recognition, people counting, pedestrian verification, etc. [0003] Existing object detection methods can be divided into traditional object detection methods and object detection methods based on convolutional neural networks. The research focus of traditional target detection methods is to skillfully design appropriate features and powerful classifiers, such as: HoG+SVM, HoG+DPM, DOT+RF, etc. Due to the w...

AI Technical Summary

Problems solved by technology

[0008] In view of the above-mentioned shortcomings in the prior art, the pedestrian detection method based on multi-scale convolution features provided by the present invention solves the defect that the existing pedestrian detection method cannot recognize small target pedestrians in the image or the recognition rate is not high

Images