Human head detection method integrating image preprocessing and deep learning target detection

Abstract

The invention discloses a quick and accurate human head detection method integrating image preprocessing and deep learning target detection. The method comprises the following steps: acquiring monitoring image data in real time; preprocessing the obtained monitoring image data; inputting the preprocessed monitoring image data into a pre-trained neural network to obtain a to-be-selected area angerin the monitoring image data and a corresponding offset value and confidence; according to the confidence coefficient of the anger output by the neural network, selecting the anger with a target; calculating the position and the size of a prediction bounding box corresponding to the corresponding target according to the offset value of the anger output by the neural network; drawing the predictionbounding box in the monitoring image data according to the position and the size of the prediction bounding box to obtain a result image; and outputting a result image and the number of heads detected in the image, wherein the number of heads is the number of predicted bounding boxes in the image. By utilizing the human head detection method provided by the invention, the quick accuracy of humanhead detection in public places can be improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a fast and accurate human head detection method which combines image preprocessing and deep learning target detection. Background technique [0002] In a large number of public places, such as large shopping malls, supermarkets, tourist attractions, large and small transportation hubs, banks, subways, and schools, it is necessary to conduct real-time analysis of the population density in surveillance cameras to ensure the orderly and stable operation of public places. For example, the real-time statistical analysis of the number of people in large shopping malls is convenient for decision makers to timely guide overcrowded areas and prevent stampede accidents. At the same time, this technology can also be applied in the campus scene to monitor and analyze the number of people in each classroom in real time. For students, they can find a suitable self-study classroom more ...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that the performance is poor in the case of mutual occlusion or insufficient light, and the more crowded the crowd, the greater the possibility of mutual occlusion, resulting in limited use of the algorithm

In addition, these target detection algorithms are not optimized for the task of "head detection", resulting in high false positive and missed detection rates when the algorithm detects low-resolution targets

[0004] And due to the limitations of the hardware conditions of the current monitoring equipment, whether it is a network camera or a wired camera, the monitoring video images are generally blurred and the noise is unstable.

Images