Research And Application Of High-speed Vision Image Processing Algorithm For Mobile Robot

As a robot integrating environment perception,decision planning and behavior control,mobile robot has certain autonomy and good environmental adaptability,and is widely used in production and life.Visual navigation technology is a key technology for the researchand development of robots,and it is an important guide and basis for robots to accurately perform established tasks.With the rapid development of information technology,it presents higher technical challenges to the speed of robot information acquisition and processing.Aiming at the problem that the existing image processing algorithms cannot meet the mobile robot's demand for real-time online processing of large amounts of data.This paper proposes a set of high-speed vision image processing algorithms for mobile robots.The introduction is as follows:(1)Aiming at the low efficiency of the existing image preprocessing algorithm,this paper proposes an image preprocessing algorithm based on FPGA.The algorithm is divided into two parts: image noise removal and image distortion correction.Aiming at the characteristics of traditional filtering methods,such as single object orientation and low operational efficiency,an Adaptive Median-Gaussian Filtering(AMGF)algorithm is proposed in this paper,which has a better suppression effect on combined noise and a faster processing speed than traditional processing methods.Aiming at the problem that the traditional image correction algorithm cannot meet the real-time processing requirement of mobile robot in software,this paper completes the implementation of image correction algorithm on FPGA,which improves the efficiency of the algorithm.In the application.The output dual-channel pre-processing RGB image reached the output frame rate of 640× 480@170fps.(2)To solve the problem of poor real-time performance of the existing image contrast enhancement algorithm,this paper proposes the Adaptive Contrast Enhencement Algorithm(ACEA)hardware implementation framework.The method of threshold judgment of RGB channel mean is used for white balance processing,which makes the white balance processing have better adaptability to different scenes.Power law transformation module and color space transformation module are integrated in the frame,which can adjust the image brightness and realize the operation of RGB image to gray image.In the application,the image contrast enhancement processing outputs the dual RGB image to reach the output frame rate of 640 × 480 @ 150 FPS.(3)To solve the problems of existing feature matching algorithms,such as large amount of computation,low efficiency of real-time processing of high frame rate image data and heavy burden on processor,this paper proposes a Fast multi-bit Description ORB(FMDORB)based on FPGA for high-speed image processing.the algorithm improves the efficiency of feature point extraction,and improves the matching accuracy of the algorithm for similar texture features.The design uses FPGA to implement and accelerate the algorithm to reduce the burden on the processor when processing large amounts of data in real time.Finally,the average processing speed of the image with a resolution of 640 × 480 at the operating frequency of 100 MHz of FPGA is only 6.2ms,and the output frame rate of 120 FPS is achieved in practical applications.The high-speed vision image processing algorithm for mobile robots proposed in this paper has been applied to the microminiature binocular vision unit of the robot,and the image acquisition and processing experiments of indoor and outdoor environments have been carried out.The algorithm stabilizes the dual-path processing at a rate of 640 × 480 @ 120 FPS The image and the feature point descriptor are output.