Target detection method, device and system based on local integral image processing

Abstract

The invention provides a target detection method, a device and a system based on local integral image processing. The method includes: step S11, reading in pixels in a local image zone and calculating local integral images in the local image zone, wherein the size of the local image zone is larger than or equal to that of a child window which is a minimum rectangular zone capable of detecting targets in the target detection; step S12, calculating rectangle features of each child window in the local image zone by using the local integral images, and performing target detection in each child window according to the rectangle features of each child window; and step S13, updating the local image zone, and repeating the step S11 to the step S12 for the updated local image zone until target detection for the whole image is completed. The target detection method, the device and the system can limit the data bit for representing pixels of integral images and save static random access memory (SRAM) resources for storing the integral images.

Description

technical field [0001] The present invention relates to target detection technology, in particular to a target detection method, device and system based on local integral image processing. Background technique [0002] The Haar-Like feature is a simple rectangular feature proposed by Viola et al. (see Document 1: Paul Viola, Michael Jones, Robust Real-time Object Detection [J]. International Journal of Computer Vision, 57(2), pp.137 -154,2002.), named for its similarity to Haar wavelets. Such as Figure 1a to Figure 1c As shown, the definition of the Haar-Like type feature is the difference between the gray level sum of the black rectangle and the white rectangle in the corresponding area of ​​​​the image sub-window, and the Haar-Like type feature reflects the local gray level change of the image. [0003] Haar-Like features are divided into three categories: edge features, linear features, center features (see Document 2: B., Ernst, A., Fast frontal-view face detection u...

AI Technical Summary

Problems solved by technology

Document 5 (Hiromoto M., Sugano H., et al., Partially Parallel Architecture for AdaBoost-Based Detection With Haar-Like Features[J]. IEEE Transactions on Circuits and Systems for Video Technology, 19(1), pp.41 -52, 2009), the integral image of the entire image is stored in the on-chip buffer, which requires 1M byte (Byte) storage space and takes up a lot of logic resources

Images