Number-of-people detection method for YOLO convolutional neural network

Abstract

The invention discloses a number-of-people detection method for a YOLO convolutional neural network, and the method comprises the steps: setting a library file creation unit, a feature extraction unit, a number-of-people judgment unit and a library file creation unit which is used for creating a standard library file which comprises a plurality of reference convolution features, network parametersand the corresponding number of people; and the feature extraction unit is used for receiving the video frames shot by the camera and extracting convolution features of the video frames so as to realize people number detection. According to the number-of-people detection method for the YOLO convolutional neural network, by detecting pedestrians in an image and distribution and semantic counting method attributes of the pedestrians, a semantic attribute learning method of the pedestrians in the image is used for assisting the pedestrians in detecting the pedestrians in the image, the influenceand interference of semantic attributes of the pedestrians in the image on the pedestrians are inhibited, the detection precision is improved, and meanwhile, the problem that a deep learning pedestrian counting method in a target image video detection scene is low in accuracy is solved.

Description

technical field [0001] The invention relates to the technical field of the number detection technology of a YOLO convolutional neural network combined with an optimized prior frame selection method and semantic information, and specifically relates to a number detection method of a YOLO convolutional neural network. Background technique [0002] In the current pedestrian detection algorithm, the images in the YOLOv3 target image detection network often have a high rate of missed detection. Considering that there is a certain correlation between the distribution of pedestrians in the target image network and the learning method of semantic counting attributes, And the accuracy is not high, so it is necessary to invent a method for the number detection system of the YOLO convolutional neural network based on the optimized prior frame selection method. Contents of the invention [0003] (1) Solved technical problems [0004] Aiming at the deficiencies of the prior art, the p...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method for detecting the number of people in a YOLO convolutional neural network, which has the ability to optimize the selection method of the prior frame, and more overall consider the IOU pedestrian classification in the selection of the prior frame classification. Confidence and other advantages solve the problem that in the pedestrian detection algorithm, the images in the YOLOv3 target image detection network often have a high rate of missed detection, considering the distribution of pedestrians in the target image network and the learning method of semantic counting attributes. A certain correlation, and the problem of low accuracy

Images