Moving object detection method and device based on depth learning

Abstract

The invention provides a moving object detection method based on depth learning. The method comprises that video images are input or collected; video image moving object detection is performed on video images to obtain the foreground region; the foreground region is expanded to obtain the position of the expanded foreground region and the corresponding sub-image; the sub-image is scaled to a fixedheight, and the scaled sub-image is transversely stitched to obtain the transversely stitched sub-image; the trained depth learning model is used to detect the object in the transverse mosaic image,the object detection frame is obtained, and the region mapped by the object detection frame in the video image is taken as the object detection region and output. Compared with the prior art, the invention can detect the moving object quickly and has high detection accuracy.

Description

technical field [0001] The present invention relates to image processing and video monitoring, and in particular, to a method and device for detecting moving objects based on deep learning. Background technique [0002] Object detection refers to the ability of computers and software systems to locate and identify each object in an image / scene. It has been widely used in face detection, vehicle detection, pedestrian counting, network images, security systems, and driverless cars. [0003] Traditional moving target detection methods generally use frame difference method or background subtraction, but these traditional moving target detection methods are easily affected by light, environment, etc., and the detection accuracy is low. [0004] In recent years, detection methods based on deep learning (such as SSD, Yolo, Faster RCNN, etc.) have a good detection effect on multi-targets, but in order to solve the problem of wide-scale multi-target detection, a large amount of compu...

AI Technical Summary

Problems solved by technology

[0003] Traditional moving target detection methods generally use frame difference method or background subtraction, but these traditional moving target detection methods are easily affected by light, environment, etc., and the detection accuracy is low

[0004] In recent years, detection methods based on deep learning (such as SSD, Yolo, Faster RCNN, etc.) have a good detection effect on multi-targets, but in order to solve the problem of a wide range of multi-target detection, a large amount of calculation is often required

Images