Automated plant classification

Team members: Bram van Riessen, Guus Paris (2016)

Project Roborchid

Automating the sorting of orchids

Hazeu Orchids                                  

Automating the accurate identification of topological features of organic products is indeed a difficult task. But no task is too hard for the Roborchid team.

The Roborchid sorting system

As a continuation from the SMR-1 project Roborchid, the team set its goal to optimize and attempt to commercialize the orchid sorting machine developed in the months leading up to SMR-2. With just two full-timers left, Bram van Riesssen and Guus Paris, from the initial four that started in SMR-1 it was going to be a tough assignment.

For some more historical insight in the Roborchid project, you can take a look at the article written, at the end of SMR-1, through this link.


The setup from SMR-1 gave an inspection accuracy of 96% over 300 plants but a very frequent and long downtime period. Still, these results were promising, which gave inspiration to continue with the optimization of the system.

Business case

The results of the first version of the orchid sorter were great, and the possibility to continue optimizating gave the team a chance to further improve the system. One of the first decisions this semester was discarding the pickup part of the transportation subsystem, fulfilled by the UR10 robot. Because of this decision, the focus shifted towards improving the rotation system and the inspection system, which are unique selling points of the Roborchid.

Leaving the robot-arm out of scope also gave a big advantage to the attempt to commercialize the system. Removing the robot-arm significantly lowers the cost price with over 50% and reduces the downtime, as picking up the plants comes with a lot of mistakes and was a big part of the frequent errors.

This means that the gains of implementing the sorting system in a production line are no longer directly replacing the two employees who are sorting all day. The direct short-term gains are limited, but the system does speed up by a large margin. The digitalization of the plant data creates long term insights in the size of the plants inside the production system.

So instead of building the business case solely around the reduction of necessary man hours, the business case focuses on significantly speeding up the sorting process. Based on the customer´s requirements, the sorting machine can modularly be upgraded with plugins.

The picture below shows the functionality tree of the sorting system. At the beginning of SMR-2 we chose to at least build the blue framed functions in the prototype. The blue boxes are the essential unctions of the machine and the orange boxes are modular functions that create value to the system but need the blue functions to operate.


The complete sorting system

The current system and prototype consists of three subsystems which are the inspection, rotation and transportation subsystems. Combined, these three subsystems are capable of sorting orchids with an accuracy of over 95%, with a speed of over 2400 plants per hour and with very little down-time.

Rotation subsystem

The first version of the rotation subsystem was designed during the SMR1 minor. The subsystem consisted out of a friction strip, which uses gravitation to rotate the plant for 360°, whilst being transported with the help of a product carrier. This solution was great because of its simplicity, but the failure rate was too high to be effective.

3d rotation system

The new design uses an electrically driven belt in combination with the friction strip. The friction strip and belt applies pressure to the pot and rotates the plants with the same speed as the product carrier is moving along the conveyer belt.

Inspection subsystem

The inspection subsystem consists of a Cognex camera vision system that checks the rotating plants for various parameters. It gives a signal to the Siemens PLC that controls the logic between all electronic and pneumatic components.  A new feature of the inspection subsystem is the implementation of network communication. All data (pictures and data strings) that are gathered by the inspection subsystem are now stored on an external drive.

Inside lightbox

This creates the possibility to compare long term plant size data. This gives clients the opportunity to compare what kind of implications other changes in the production line have on the size of the plants.

Transport subsystem

The transport subsystem is responsible for physically moving the plants through the system. It includes the conveyer belt and the components that regulate the plant flow over the conveyer belt. This is done with pneumatic cylinders, which are controlled by a Siemens Logo! PLC. The Siemens Logo! PLC signals the pneumatic switches based on the signals it receives from a magnetic sensor and the 2D-Cognex camera.

Test results

Over the course of the last two weeks of the project, the upgraded prototype was transported to Hazeu orchids to do further testing and measure the specifications of the machine. An initial test was executed to analyze more than 350 plants to calibrate the system and make some changes where necessary. Later that week, a full simulation was run with over 650 plants, about 1% of Hazeu´s weekly production. The division leader of Hazeu was present during this test to cooperatively determine the success-rate of the machine.

During testing the inspection system made 12 inspection errors and between 15 and 20 rotational errors. These numbers led to an accuracy of more than 95%.


A big chunk of the time spent on this project went into efforts to commercialize the sorting machine. This means that the team had lots of meetings with different people about the commercial future of the sorting machine on a regular basis. These meetings were key in setting up a successful business case. Finally these efforts led to a presentation and demo with over eight entrepreneurs and two large-scale plant nursery owners.

All attendees showed enthusiasm and some showed serious interest in investment.

Beyond SMR-2

After testing and working long days on the sorting system, the team built a foundation for looking towards the future. During the SMR-2 minor the team was able to significantly improve the sorting system and is confident of pushing the current prototype to its maximum performance. Already the team has identified numerous ways of optimizing system components. This creates confidence to move ahead with the project.

With the available components and support from SMR-minor coordinators T. Brilleman and F. de Wit, the team was able to develop a system that can have commercial success.

Team Roborchid,

Bram van Riessen, Guus Paris, Rowan Overdevest