Students: Sander Bal – Simon Navarro – Simon Boerwinkel – Marc Ziedses des Plantes
Pot Chrysanthemum and the gebr. Nederpel
In Europe, we love the chrysanthemum, but not only in Europe. You can buy the plants all over the world. Nowhere the plant is so popular as in Japan. Kiku, such as the chrysanthemum is called there, can be found on the imperial weapon, the passport and the 50-yen coin. This is because people in Japan think that the chrysanthemum brings luck. The worldwide popularity explains why the Nederpel greenhouse produces 8,5 million pot chrysanthemums per year. You can imagine when some processes at the greenhouse are still done by humans it takes a lot of effort and money. This is why the greenhouse wants to automate some of the processes.
Figure 1: Full grown Chrysanthemum
In the current situation, the pots are automatically filled with earth and placed in tables of approximately two by four meters. After eighteen days the plants are big enough to be moved off the tables and placed in another part of the greenhouses. Before the plants can be moved, they need to get “pinched”. Pinching is taking off the top part of the plants. This ensures that the plant will grow better and more to the sides instead of straight up, which ensures a better looking plant. At the moment five persons are pinching the plants by hand. They pinch approximately two tables an hour, which means that they pinch between 70 and 117 plants per minute. The pinched tops are discarded into a large basket which gets emptied into a waste container when it is full.
Figure 2 : Plants are Pinched on these tables
Goal of this Project
The main goal was to make a system that can automatically top one chrysanthemum in a pot without the assistance of a human. The secondary goal was to scale up the system so it’s capable to cut three plants in a pot.
The requirements we had to take into account for this project:
- After pinching, the plants need to have four to seven leaves.
- All the plants in a pot need to have approximately the same height.
- The pinched tops need to be removed and discarded.
- Leaves can’t be damaged (this will add one week to the total grow time).
- The cut/pinch doesn’t have to be “clean”.
The complete solution consists of four subsystems:
- The transport system;
- The detection system;
- The cutting and discarding system;
- The operating system.
Figure 3: System overview
The transport system consists of the conveyor belt and the turning/holding mechanism. The conveyor belt (1) transports the pots into the cutting area. When the pot moves in front of the turning/holding mechanism (2), a sensor (3) gives the signal to the valve so the pneumatic cylinder (4) will extend. The pot turns in front of the camera so the plants can be detected.
Figure 4: Transport system
The detection system uses a webcam (5) to send an image to a computer. Specially written software, written in Halcon, is used to recognize the plants and determine the cutting position. To ensure the detection of the plant the robot enclosure has been covered to keep as much light out as possible. With the use of a spotlight, the plant is highlighted which makes it easier for the software to detect it. Coordinates are interpreted by the software and send to the robot via an Ethernet connection.
Figure 5: Camera setup Figure 6: Detected cutting position
When a plant is detected the turning mechanism stops and holds the plant in place so the robot can cut the plant. For the cutting, an End of Arm Tool has been designed that keeps hold of the cut plant part so it can be moved to the backside of the conveyor, where it is dropped. After cutting all the plants in the pot, the turning/holding mechanism is deactivated and the conveyor transports the pot out of the robot area where a worker can pick it up. The waste is deposited in a bin at the end of the conveyor.
Figure 7: End Of Arm Tool (EOAT)
Figure 8: Pinched plant and the discarded top
An operating system is used to make everything work together with at its base a Programmable Logic Controller (PLC).
Figure 9: Information stream diagram
Major decisions during the project
The first idea was to make a vision program that was able to count the leaves of the plant and afterward give the coordinates of the best cutting spot. After some research, this proved to be impossible, so statistics were needed. This statistic will give a good analysis of the average height per leaf. Through the statistics, a region of interest was chosen and used in the vision program.
Figure 10: Leaf height statistics
The university used Cognex software for the vision part. This program proved to be inadequate for organically use, so another vision program was needed. Halcon appeared to be sufficient for this project. However, this software can’t work with the Cognex camera’s from school, so a normal webcam was necessary.
The top of the plant must be pinched but also discarded, to do this with two separate systems appeared to be inconvenient. After a lot of testing, the group found a solution to make an End of Arm Tool that was capable of doing both things. It will cut the top and hold it until the robot moved away from the plant, then it will drop it at the other side of the conveyor. This will guide it into a trash can.
The main goal of precision cutting one plant with the use of a vision system and the robot has been achieved. Due to time restrictions, it was not possible for the group to address the secondary goal of cutting three or more plants in one pot.
The original approach of using vision to determine the place of the stem and every leaf on the plant was abandoned in favor of a smaller detection region that was determined using statistics.
Although the final setup is used for cutting one plant, it is scalable with the implementation of a working turning mechanism. Due to the time restriction of seven weeks, it was impossible for the group to tackle this problem successfully.
Though the assignment was tough, the group has sufficiently scaled it down during the start-up of the project. With hard work and long days, the group has made a working proof of concept within the set time limit.
This solution is a beginning and there must change a lot in the grower factory before this can be implemented. But the goal of this project was to come up with an idea to pinch a plant by using a robot. We found a solution to find a cutting location and send it to a robot that will cut it there