深度学习算法下的采摘机器人系统优化研究OA
Study on System Optimization for Picking Robot Based on Deep Learning Algorithm
为了实现苹果采摘的自动化,基于图像深度学习算法,对采摘机器人系统进行优化设计.首先,采用高斯滤波的方法,对摄像头采集的图像进行滤波处理;其次,在 VGG16 深度学习网络算法的基础上,增加区域推荐学习网络,进而减少识别样本容量,提高识别率;再次,采用单目视觉系统和激光测距器,实现目标苹果的图像坐标向实际空间坐标的转化,并采用双反馈系统实现采摘机械臂控制;最后,对系统进行测试.结果表明:系统具有良好的目标识别精度和机械臂控制精度.
In order to achieve the automation for apple picking,this optimization for picking robot was designed based on image depth learning algorithm.Firstly,the filter processing was done for image collected by the CCD camera,based on gaussian filtering.Secondly,regional recommendation learning network was added to VGG16 deep learning network,so that identification sample size was reduced,and recognition rate was improved.Thirdly,coordinate transformation be-tween image coordinates and actual spatial coordinates was achieved by monocular vision system and laser rangefinder,picking manipulator control was achieved by double feedback system.In the end,the system was tested,the result showed that this system had accuracy target recognition and manipulator control accuracy.
张军凯;李欣;韩俊先;赵娟;程龙雪
河北机电职业技术学院 电气工程系, 河北 邢台 054000
农业工程
采摘机器人改进型深度学习网络高斯滤波视觉定位
picking robotimproved deep learning networkgaussian filteringvisual positioning
《农机化研究》 2024 (004)
58-62 / 5
河北省高等学校科学技术研究项目(ZC2021222)
评论