| Machine vision has developed rapidly in recent years and has attracted the attention of researchers from various countries.Vision-based handwritten digit recognition and target tracking technology has always been a hot topic in the field of machine vision.In this paper,based on the RoboMaster competition,under the condition of the embedded microprocessor NUC7i7 BNH,the image is collected by the monocular camera,and the captured image is processed to complete the recognition of the handwritten digits and the tracking of the single target.The calibration of the USB camera in the image acquisition system is carried out by using the checkerboard calibration method,and the acquired image is grayed,filtered,enhanced contrast and image binarized.Among them,the maximum entropy method is used to binarize the image,and according to the particularity of the digital panel,it is improved to obtain a clear binarized image.The classic target location method is improved by studying the characteristics of the classical target localization algorithm and the digital panel.A target localization algorithm based on the combination of edge and geometric relations is proposed.The accuracy of the algorithm for target recognition and location is verified by experiments.The method of support vector machine is used to identify handwritten digits,and the multi-classification problem is solved by one-to-one training method for samples,and cross-validation is performed by 5-CV.Different kernel functions were compared and tested,and the radial basis kernel function was used,and the parameters were modified to improve the recognition rate.Based on the traditional color recognition algorithm,a method based on improved color recognition and threshold segmentation is proposed to detect and identify moving targets.In the STM32 MCU,the C language programming control program is used,and the PID algorithm is used to control the motor to complete the tracking task of the target. |
PDF Full Text Request