Yolov3-based personnel target detection method

Abstract

The invention relates to a Yolov3-based personnel target detection method, which comprises the steps of obtaining an image, and setting Anchor parameters by using a K-Means algorithm when a Yolov3-based reference network is constructed; taking a Darknet-53 network as a backbone network; introducing a feature pyramid structure to perform feature extraction on the multi-scale target; calculating theloss of the prediction frame offset by using a cross entropy loss function; designing the scale of the Anchor according to the height-width ratio of the personnel target; replacing the Darknet-53 network with a MobileNet_v2 network; improving the feature pyramid structure by introducing hole convolution; and after-treatment optimization is carried out by introducing an IoU confidence coefficientand a soft-NMS algorithm, obtaining an improved Yolov3 network, and identifying and detecting a personnel target. Through optimization and improvement of the invention, faster and more accurate detection of personnel targets can be realized.

Description

technical field [0001] The invention relates to the technical field of computer vision applications, in particular to a Yolov3-based personnel detection method. Background technique [0002] Traditional object detection algorithms include background difference method and frame difference method, but these methods are easily affected by factors such as lighting and complex textures, resulting in poor final detection results. The advanced features learned by the deep convolutional network in the image have better robustness. This advanced feature is not affected by the illumination but is expressed as the outline and texture information of the target itself. Therefore, more and more researches The latter chose to use a convolutional neural network-based approach for object detection and recognition. In 2014, R. Girshick and others first proposed a target detection method based on feature regions. Since then, target detection algorithms based on convolutional neural networks h...

AI Technical Summary

Problems solved by technology

W. Liu et al. proposed the SSD target detection algorithm in 2015. It is the second one-stage target detection algorithm in the field of deep learning. The main contribution of the SSD algorithm is that it uses deep and shallow feature layers to predict the detection frame of the target at the same time, and Using multi-resolution technology to detect multi-scale targets, the SSD algorithm has significantly improved the detection accuracy of small targets compared with the Yolo algorithm, but its average accuracy is still lower than the Two-stage target detection algorithm

Images