| Under the present conditions of our country,manual sorting is still the major means for garbage sorting which is of many shortcomings,such as poor working environment,large labor intensity,low sorting efficiency and low automation degree,cannot meet the recycling of environmental resources and social progress in our country.With the improvement of industrialization in our country,automatic systems are widely applied in many industries.However,lacking of automatic sorting is the current situation due to started late,less investment,and technical difficult.Therefore,it is of great significance to improve the intelligent sorting system of solid waste.Firstly,the design of the sorting system for solid waste is presented.The sorting system is consisted of the unit of target recognition and the unit of sorting,which lay the foundation for the following designs.The angle and location information of the target object are calculated by recognition algorithm,and manual works are replaced by industrial robots.Secondly,an improved algorithm based on Faster-RCNN is put forward to predict the position and posture of objects in the complex solid waste environment.An improved design of VGG16 network is showed in the paper to adapt to recognize targets like bottles based on Hyper-Column with human-computer interaction error correction to improve the correctness and effectiveness.By analyzing the algorithm principle and improved algorithm,hardware of the unit of target recognition is designed.Then,kinematics analysis of robots' manipulators and hardware of the unit of sorting are described.The coordinated control of multi robots is designed.Especially to ensure the high reliability of sorting system the end effectors of manipulator is designed in complex environment using.Finally,to verify the performance of improved algorithm based on FasterRCNN is test through pictures taken by lab and field,and a sorting experiment system is set up to test by simulating actual situation. |
PDF Full Text Request