Quality phenomics new ways to determine quality based on data and prediction

Transcription

1 Quality phenomics new ways to determine quality based on data and prediction Smart Horticulture Asia 2016 Hong Kong Rick van de Zedde, 8 th of September 2016 Introduction Rick van de Zedde, business developer/ senior researcher, 11 years at Wageningen University and Research. Background: Artificial Intelligence. Focus: computer vision/ robotics Aim of this presentation: inspire you with examples on data collection for quality measurements from research perspective 1

2 Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective Wageningen University & Research 2 organisations: Wageningen University & 9 contract research institutes Turnover > 700 mln 6,000 employees 10,000 students > 100 countries! China office: Zhen Liu zhen.liu@wur.nl Wageningen UR - campus 2

3 Wageningen University & Research Location: Europe The Netherlands Wageningen Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective 3

4 Robotics within Wageningen UR Wageningen UR - profile Wageningen University FBR: a CRO Industry Fundamental research assessment Applied Research assesment Demonstration assesment Implementation business impact Training Pilots Consultancy Knowledge Push Market Pull 4

5 Example: PlantSampler robot Video by: Piet Oomen (IsoGroup)/ Rick vd Zedde (Wageningen UR) Under development: Robot sampling for DNA analysis 5

6 6

7 Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective Definition of Phenomics en link to quality Phenomics The measurement of phenomes the physical and biochemical traits of organisms as they change in response to genetic mutation and environmental influences. Goal To gain significant progress in improving the quality and productivity of crops, we need to understand and exploit the genotype and the phenotype of plants/ products in a changing environment (climate change/ diseases). 14 7

8 <PhenomicsNL plaatje hier> 15 Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective 8

9 Example case: Mango quality issues Huge variability on the tree Huge variability in the supermarket Ripening of (sub) tropical fruit mango, avocado, banana,... Ready-to-Eat fruit in supermarket show huge variability How can we ripen more accurate and efficient? Measure of initial ripeness and other product characteristics on individual fruit Understand quality development Apply customized ripening protocols 9

10 Wageningen approach Quality Chain development conditions + Initial Quality + = Predicted Quality Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective 10

11 Quality phenomics- avocados Firmness: important quality aspect Avocado ripening Acoustic measurement Non-destructive Firmness: Acoustic measurement Avocado: Firmness: Acoustic versus fruit texture analyser Different mango batches: 22 11

12 Time lapse & automated data extraction Example: rot area development of Penicillium expansum inoculated Elstar apples during shelf life 23 12

13 Sensor/ probe on robot arm 25 Measurements and results pears % brown spots on skin % brown spots on skin Colour 26 13

14 Feasibility results: brix in tomato Brix measurement in 20ms using a self-made NIR probe and mount it on robot arm for automation. Self made NIR probe 27 Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective 14

15 EU project PicknPack background Food processing factories have to be flexible: Large number of products and packaging variations Small batches and retailers place theirs orders late To meet increasing demands: Manual labour, because people are flexible Robots/machines are not flexible (yet) EU-project: PicknPack 30 15

16 Principle image analysis approach Step 1: Segment fruit, stalk and background Step 2: Individualize tomatoes Step 3: Analyse each single tomato and stalk for defects Tomato analyses Colour (ripeness) Cracks/cuts Blossom-end rot Weight Stalk analyses Yellow calyx Mold on calyx Mildew Sensor fusion - Stem/tomato segmentation Segmentation based on colour and 3D info 16

17 Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective EasyFlow automated check-out Concept: Supermarket check-out with no barcodes. Identify 30,000 different products. Solution machine learning on multiple sensors: computer vision, NIR technology, weight and statistics. Continuous self learning shared with other Easyflow check-outs Result Hitrate 99%, better than a human cashier. 17

18 ITAB Match X project Nicole Koenderink (PL), Don Willems, Roeland vbatenburg, ea. Outline presentation Wageningen UR Agro Food Robotics Quality Phenomics Example case Automated quality measurements Packaging and robotics Automation in retail Future perspective 18

19 Future work Quality Phenomics Collect reliable data in the chain using automation: Define and exploit relation non-destructive robot measurement and manual/ destructive lab measurements. Vision-guided robotics for dedicated measurement on specific (flat) surfaces of individual objects. Add gripper to robot for exploiting sensors like firmness, spectrometers, colour, etc. Understand and predict quality developments! Mission Our research groups is developing sensor/ robot technology for non-destructive quality inspection and data collection of fruits and vegetables. Industrial partners are invited to join! 38 19

20 Acknowledgement Colleagues at Wageningen UR: EU projects: Henry Boerrigter Erik Pekkeriet Gert Kootstra Plus industrial partners: Gerwoud Otten Franck Golbach IsoGroup (P. Oomen): Eelke Westra Freek Daniels WPK (E. vd Arend): Ernst Woltering ITAB (J. Möller): Jan Verschoor Fatima Pereira da Silva Questions? Thank you for your attention! Rick van de Zedde Zhen Liu (China office) 40 20