Real-time human head detection method based on stair scene

Abstract

The invention discloses a real-time human head detection method based on a stair scene in the field of computer vision. The real-time human head detection method comprises the following specific steps: S1, collecting a large number of picture data sets of the stair scene; S2, labeling the data set, wherein a labeling box needs to contain head and shoulder information; S3, segmenting the data set into a training set, a test set and a verification set; S4, enhancing the training set data; S5, extracting annotation box data in the training set, performing kmeans clustering on annotation box information, and selecting categories to obtain different anger information; S6, an FCHD + FPN network architecture is constructed; S7, according to the marked stair head training set, training by using anFCHD + FPN network model; S8, testing the accuracy of the trained model in the verification set; and S9, using the generated model to detect a human head in a real stair scene, selecting an anger through a clustering method based on a real-time human head detection method of the stair scene, combining shoulder information to adjust a labeling area, Meanwhile, fusing an FPN network to improve an FCHD method to improve the detection accuracy.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a real-time human head detection method based on a staircase scene. Background technique [0002] The existing head detection method has two ideas, one uses the regression algorithm idea, and obtains the crowd density heat map according to the image regression, this method can only show the crowding of the flow of people, and cannot locate the specific location of the person At the same time, the resolution of the image is very high; another method using target detection, such as SSD, Yolo, Faster-rcnn series algorithms, to detect the number of people, these algorithms are less effective in the case of mutual occlusion, And it is difficult to meet the requirements simultaneously on the accuracy and speed of detection. [0003] FCHD is the latest detection algorithm for the scene of head detection, but the anchor of FCHD only chooses two sizes. In actual use, the generali...

AI Technical Summary

Problems solved by technology

[0002] The existing head detection method has two ideas, one uses the regression algorithm idea, and obtains the crowd density heat map according to the image regression, this method can only show the crowding of the flow of people, and cannot locate the specific location of the person At the same time, the resolution of the image is very high; another method using target detection, such as SSD, Yolo, Faster-rcnn series algorithms, to detect the number of people, these algorithms are less effective in the case of mutual occlusion, And it is difficult to meet the requirements of detection accuracy and speed at the same time

[0003] FCHD is the latest detection algorithm for the scene of head detection, but the anchor of FCHD only chooses two sizes. In actual use, the generalization ability is not good, and the missed detection rate is high, because the size of the head is related to the position of the camera and the human body. The distance relationship is relatively large

Images