In order to improve the recognition accuracy of fruit and vegetable robot picking,SDD network is used for real-time detection.The structure of the basic network and the default box of the detection layer are improved,the number of convolution layers is increased,and the calculation parame-ters are reduced,so that it is convenient to recognize fruits and vegetables of different scales,and an au-tomatic target recognition framework based on the improved SDD model is obtained.The activation function is analyzed,and it is found that the Relu function has faster convergence speed.The activation operation is carried out by using the Relu function,and the training results are analyzed.The results show that the improved SDD model has faster convergence speed.When the confidence is 0.45,the im-proved SDD model has ideal accuracy,recall,FI value and IOU value,which can meet the recognition requirements of fruit and vegetable robot picking.
关键词
果蔬/采摘机器人/改进SDD模型/自动化识别
Key words
fruits and vegetables/picking robot/improve SDD model/automatic identification
维普
万方数据