Robot vision system and processing method thereof

Abstract

The invention discloses a robot vision system. The robot vision system comprises an image acquisition mechanism and an image processing mechanism. The image processing mechanism comprises a host processing module and at least one slave processing module in communication with the host processing module via a server. According to the robot vision system adopting the structure, the host and the at least one slave communicate through the cloud server, all the high-performance calculation and large storage modules of the robot can be unloaded, and all operation modules, storage modules and algorithm modules perform by virtue of a remote cloud platform, so that the cost of the robot is reduced, and the robot is lighter.

Description

technical field [0001] The invention relates to a robot technology, in particular to a robot vision system and a processing method thereof. Background technique [0002] With the development of science and technology, robot vision technology has been greatly developed, and the corresponding visual recognition algorithm has gradually matured. However, problems such as low sampling rate, large amount of image data, high algorithm complexity, and long running cycle of the robot vision system hinder the popularization and application of robot vision. With the continuous maturity of cloud technology, the concept of cloud robot has emerged. Cloud robot is the combination of cloud computing and robotics. Just like other network terminals, the robot itself does not need to store all data information or have super computing power, but can connect to relevant servers and obtain the required information when needed. This method is an important way to solve the above bottleneck proble...

AI Technical Summary

Problems solved by technology

[0003] (1) The traditional robot body carries a large amount of computing and storage equipment, which leads to the contradiction between the function of the robot and the lightweight and low cost of the robot itself;

[0004] (2) The traditional robot visual recognition platform adopts the stand-alone training mode of high-performance computers, which leads to the inability to reuse the system code and it is difficult to solve the code usage problem of heterogeneous robots;

[0005] (3) Traditional parallel computing mostly relies on high-performance computer clusters. High-performance computers are replaced too quickly, resulting in high costs and a relatively fixed number of server nodes, which is not flexible enough. Traditional private image recognition platforms lack relevant network security protection, which is very easy Cause hackers, viruses and other interference, and platform server hardware maintenance and management, waste a lot of manpower and material resources

[0006] (4) At present, the robot does not have a unified system structure, and the modularization and standardization of its functional components lacks corresponding standard systems and interface specifications. It can be reused and replaced, so there is no specialized support and industrial chain division of labor, and industrial-level production cannot be realized, resulting in high manufacturing costs, which restricts the formation and development of its industry

[0007] The above deficiencies lead to problems such as low image recognition efficiency, long cycle, and inconsistent data structure and protocol of robots, which hinder the application and promotion of vision in the direction of robots.