Embedded real-time human shape detection method and device

Abstract

The invention relates to the field of computer vision, in particular to an embedded real-time human shape detection method and device, and the method comprises the steps: obtaining a human shape detection model based on a convolutional neural network through training in advance; carrying out image acquisition in a to-be-detected space range by utilizing a human shape detection device, and calculating an ROI region; selecting a detection stage according to the ROI area, dynamically adjusting the image resolution, and inputting the adjusted image into a human shape detection model; carrying outconvolution operation on the input image by utilizing a human shape detection model and combining a multi-core parallel algorithm and a fast convolution algorithm, and extracting feature data; and obtaining a plurality of human-shaped candidate frames according to the feature data, screening and fusing the human-shaped candidate frames, and finally determining a human-shaped target. Model trainingis carried out based on the convolutional neural network, the detection precision is high, meanwhile, the multi-core parallel and fast convolution algorithm is combined, the model resolution and focal length are dynamically adjusted, the hardware requirement is low, and the miniaturization requirement of a human body detection module can be met.

Description

【Technical field】 [0001] The present invention relates to the field of computer vision, more specifically, to an embedded real-time humanoid detection method and device. 【Background technique】 [0002] Humanoid detection is an important research topic in computer vision, and it is widely used in intelligent video surveillance, vehicle assisted driving, intelligent transportation, intelligent robots and other fields. The mainstream humanoid detection methods are divided into statistical learning methods based on artificial image features and deep learning methods based on artificial neural networks. The logarithmic average missed detection rate of the statistical learning method for the human target is generally about 15%, and the speed is fast, and real-time detection can be performed on the CPU, but it is easily affected by occlusion and background. The logarithmic average missed detection rate of the deep learning method is as low as 7%, and it has strong anti-interferenc...

AI Technical Summary

Problems solved by technology

The logarithmic average missed detection rate of the statistical learning method for human objects is generally about 15%, and the speed is fast, and real-time detection can be performed on the CPU, but it is easily affected by occlusion and background

The logarithmic average missed detection rate of the deep learning method is as low as 7%, and it has strong anti-interference ability and can better distinguish between occlusion and background, but the running speed is slow, which is not conducive to realizing real-time detection on embedded systems with limited hardware resources. detection

[0003] It can be seen that there is a serious contradiction between accuracy and performance in the existing human detection methods. When the detection speed is fast, the detection accuracy is low; and if the detection accuracy is high, the hardware performance requirements are high, and it is difficult to meet the small size of the human detection module. the question of requirements

Images