Processing method for front-end identification of edge computing box

Abstract

The invention discloses a processing method for front-end identification of an edge computing box. A Huawei Hisilicon HI3559AV100 processor is used as a carrier of edge computing box processing algorithm identification and a series of tasks. Image algorithm identification is carried out at the upper end instead of the rear end; the method is characterized in that an HI3559AV100 processor core of the edge computing box is composed of a dual-core ARM Cortex-A73 at 1.6 GHz, a dual-core ARM Cortex-A53 at 1.2 GHz and a single-core ARM Cortex-A53 at 1.2 GHz. According to the invention, the identified algorithm program is installed at the identification site, so that algorithm identification is achieved at the occurrence site without depending on the network.

Description

technical field [0001] The invention relates to the technical field of front-end identification, in particular to a processing method for front-end identification of an edge computing box. Background technique [0002] There are already many identification-based products on the market. For example, mobile phones have some recognition functions, but the functions of mobile phones are not single. At the same time, the processing of mobile phone image recognition depends on the performance of its processor, and it is difficult for low-end mobile phones to process smoothly. For complex image processing tasks, since it is not a product with a single function, it will not only optimize image processing and ignore other functions, so the optimization ratio is balanced, and single-function recognition products, such as face tracking and recognition at stations The recognition of mobile vehicles on the road is based on a single function product, and the requirements for the front-end...

AI Technical Summary

Problems solved by technology

[0002] There are already many identification-based products on the market. For example, mobile phones have some recognition functions, but the functions of mobile phones are not single. At the same time, the processing of mobile phone image recognition depends on the performance of its processor, and it is difficult for low-end mobile phones to process smoothly. For complex image processing tasks, since it is not a product with a single function, it will not only optimize image processing and ignore other functions, so the optimization ratio is balanced, and single-function recognition products, such as face tracking and recognition at stations The recognition of mobile vehicles on the road is based on a single function product, and the requirements for the front-end hardware are not high. The purpose is to reduce the cost of the front-end layout. Cameras that can provide corresponding image quality can meet the requirements. On the remote server, the recognition is performed by the program in the server, relying on the computing power of the server, and the server provides multi-task or even distributed image recognition, which requires stability in the recognition scene (that is, the space where the recognition object is located). Large-bandwidth network transmission, so it will cost a lot to ensure the on-site network

For relatively remote identification sites, it is difficult to provide large bandwidth and stable transmission rate. For example, some construction sites cannot transmit through wired networks, but can only be based on wireless networks. The transmission efficiency of wireless networks depends on the network infrastructure of local operators. Development

[0003] Since the core of the technology has been moved to the site, the conditions on the site have become harsh, and artificial damage to hardware facilities will bring more losses

Front-end hardware devices are more vulnerable to hacker attacks due to the lack of comprehensive security mechanisms

Images