Hardware design method of adaboost face detection algorithm based on haar characteristics

Abstract

The invention discloses a hardware design method of an adaboost face detection algorithm based on haar characteristics. The device comprises a to-be-detected image integral graph column-by-column reading and loading control unit, a cascade strong classifier integral graph data storage unit, a Haar characteristic parameter coding and storage unit and an integral graph data processing and detectingunit. The to-be-detected image integral image only maintains 32 * 32 fixed window integral data in a detection sub-window. An integral graph data storage unit which reads from an off-chip DDR sequenceand writes the integral graph data storage unit into two detection sub-windows formed by 33 groups of 28X32RAMs is loaded according to columns. It is guaranteed that the integral graph data processing and detecting unit of the series-parallel mixed assembly line processing architecture is in high-speed detection continuously, occupied hardware resources are small, and on-chip storage space is saved. The Haar characteristic parameters are fixed, and the detection algorithm is realized by a fixed point arithmetic unit, so that the calculation complexity is reduced, and the face detection precision is not reduced.

Description

technical field [0001] The invention relates to a hardware design method of an adaboost human face detection algorithm based on haar features. Background technique [0002] Face detection has been widely used in the fields of the new generation of human-computer interaction, intelligent security, intelligent monitoring and image retrieval. With the development of embedded technology and smart devices, many face detection application platforms have developed from PCs to portable smart devices based on embedded platforms. Due to the large amount of data that face detection algorithms need to access and the high computational complexity, currently Only high-performance embedded devices can meet the performance requirements of real-time detection, and the system cost is relatively expensive. However, the computing power of low-cost embedded platforms is quite limited, and real-time face detection of embedded pure software cannot be realized. Therefore, it is often necessary Con...

AI Technical Summary

Problems solved by technology

With the development of embedded technology and smart devices, many face detection application platforms have developed from PCs to portable smart devices based on embedded platforms. Due to the large amount of data that face detection algorithms need to access and the high computational complexity, currently Only high-performance embedded devices can meet the performance requirements of real-time detection, and the system cost is relatively expensive. However, the computing power of low-cost embedded platforms is quite limited, and real-time face detection of embedded pure software cannot be realized. Therefore, it is often necessary Consider adding a hardware acceleration module to make up for the lack of computing power of the embedded platform, and how to design a low-cost, high-speed, real-time hardware structure that meets the characteristics of the face detection algorithm has become the core technical problem to solve the embedded face detection

Images