Embedded type visual servo control method based on multi-feature fusion

Abstract

The invention discloses an embedded type visual servo control method based on multi-feature fusion, and the method comprises the following steps: building a core image processing module through an FPGA, processing the collected image through the core image processing module, and extracting the geometric features; 2, building an embedded type parallel processor based on the FPGA in the CGA system, integrating with the features of a plurality of types of geometric images through the embedded type parallel processor, and building a target posture relation mathematic model; 3, carrying out the parameterization and parameter optimization through the built mathematic model, obtaining a fusion mathematic model approximate linearization function, and solving the posture parameters of a target relative to a camera; 4, carrying out the decomposition of position and posture in the CGA system, and carrying out the visual servo control. According to the invention, the method solves problems of poor noise sensitivity, poor fault tolerance, the capability of just being adapted to a plane target, local minimum and bigger calculation amount. Moreover, the method is low in power consumption, is low in cost, is high in calculation efficiency, and is good in universality.

Description

technical field [0001] The invention relates to the technical field of visual servo control of robot vision systems, in particular to a multi-feature fusion method for creating embedded visual servo. Background technique [0002] Visual servo control uses visual information as feedback, has a larger amount of information, improves the flexibility and accuracy of the robot system, and plays an irreplaceable role in robot control. At present, the feedback of visual servoing mainly has the form based on position, image feature and multi-view geometry. For the commonly used image feature method, a single image feature is used as visual feedback for servo control, which has many problems such as sensitivity to noise, poor error tolerance, only suitable for planar targets, local minimum and large amount of calculation. The fusion of multiple image features can process and synthesize information at multiple levels, levels, and aspects. Therefore, based on the visual feedback info...

AI Technical Summary

Problems solved by technology

First of all, no matter the hardware platform is FPGA or ASIC, in the case of using standard mathematical methods (especially floating-point calculations), there will be obvious problems in the calculation efficiency of complex geometric functions, which leads to the realization of algebraic calculations (especially multiplication, addition and accumulation). Operation) The number of hardware operation blocks required is large and the structure complexity is high

Secondly, in order to effectively solve problems such as complex multi-dimensional scientific computing and advanced graphics applications, the structure of the hardware processing system will also be too complicated

It can be seen that the traditional algebraic method has too high requirements on hardware when dealing with 3D and above problems, including the complexity and number of hardware processing blocks themselves.

Images