This work is a kiwifruit picking robot system based on target recognition technology, using STM32F407ZGT6 as the main control.
1) Efficient recognition and positioning: The system uses the YOLOv5 target detection model and combines it with the attention mechanism
The SGBM stereo matching algorithm can quickly and accurately identify and locate the spatial position information of kiwifruit. This enables the robot to navigate and pick accurately;
2) Flexible gripper design: The robot is equipped with a flexible gripper that integrates pressure sensor technology to
The clamping force is automatically adjusted according to the size and shape of the kiwifruit to prevent the fruit from being damaged. This feature greatly improves the success rate of picking and the integrity of the fruit;
3) Efficient actuator: The picking actuator adopts sliding module, which has
It has the advantages of low cost, simple control and wide coverage. This mechanism not only improves the overall efficiency of the system, but also realizes the integrated operation of picking and collecting, which significantly improves work efficiency;
4) Intelligent control system: The system uses Jetson Nano processor and STM32F407ZGT6 main control, combined with advanced algorithms and sensor technology, to achieve the processing of visual information, calculation of three-dimensional coordinates, and precise control of the gripper;
5) Vehicle module design: The system designs a Bluetooth wireless remote control robot chassis with four-wheel drive
The mobile function can adapt to complex terrain and improve mobility and flexibility in the orchard. This design ensures the stability and efficient operation of the system in different environments.
Main innovations
-
The overall structural design uses sliding modules to replace traditional robotic arms, simplifying the structure, reducing costs and expanding
the picking coverage;
-
Improved detection algorithm model, based on YOLOv5 and introducing attention mechanism and optimized loss function, which
improves detection accuracy and bounding box positioning accuracy;
-
The flexible gripper is equipped with a pressure sensor, which can accurately
grip fruits of different sizes through feedback control to prevent damage to the fruits.
-
The integrated binocular vision stereo matching algorithm obtains three-dimensional position information and achieves high-precision spatial positioning.
