Pedestrian detection method based on deep learning and detection device

Abstract

The invention discloses a pedestrian detection method based on deep learning and a detection device. The method comprises the steps that video data is acquired; multiple key areas where pedestrians possibly exist are positioned according to the video data; the key areas where pedestrians possibly exist are subjected to multiple iterations through a deep convolutional neural network, and pedestrian features after adjustment and filtration are obtained to judge whether pedestrians exist in the key areas; and if pedestrians exist in the key areas, a bonding frame and credibility of each pedestrian detection result are output. When the method is applied to pedestrian detection, the detection effect of the key areas and the detection effect of the pedestrians in the key areas can be improved, the purpose of high-definition video real-time pedestrian detection required by a true application scene is achieved, detection precision and detection efficiency are both improved, and the method is simple and easy to realize.

Description

technical field [0001] The invention relates to the field of computer multimedia technology, in particular to a pedestrian detection method and detection device based on deep learning. Background technique [0002] In related technologies, pedestrian detection methods using background modeling and statistical learning may achieve better pedestrian detection efficiency or accuracy under certain conditions, but neither of these two methods can meet the requirements of practical applications. Among them, the background modeling method is generally more complex, resulting in the inability to meet the needs of real-time detection in practical applications, and the method based on statistical learning is difficult to train a general-purpose pedestrian detection classifier due to the complexity of classifier training, especially when the sample size is large. And the training time of the classifier is very expensive. If some key areas in the video content can be detected in advance...

AI Technical Summary

Problems solved by technology

[0002] In related technologies, pedestrian detection methods using background modeling and statistical learning may achieve better pedestrian detection efficiency or accuracy under certain conditions, but neither of these two methods can meet the requirements of practical applications

Among them, the background modeling method is generally more complex, resulting in the inability to meet the needs of real-time detection in practical applications, and the method based on statistical learning is difficult to train a general-purpose pedestrian detection classifier due to the complexity of classifier training, especially when the sample size is large. And the training time of the classifier is very expensive. If some key areas in the video content can be detected in advance, and then the accuracy of pedestrian detection in these key areas can be improved, it will be able to improve both time efficiency and detection accuracy.

Images