Dallas Kroon in our group has written a phenotyping application for Android smart phones. Data is collected by use of a BlueTooth barcode scanner and barcode-based tools. This new system creates new opportunites to improve phenotyping. Because it is network-aware, data can be immediately stored in a centralized, remote data repository. Additionally, if the entity being measured has been measured previously, variance limits can be determined and the user can be asked to verify outlier measurements, thereby providing additional confidence in the data. It can also help to minimize missing data by directing phenotypers to as-yet unmeasured plots.
This system collects GPS information could both alleviate the need to scan plot tags as well as allowing much improved field correction of environment variability. Some of the requirements for this system are as follows: 1) lightweight enough to be carried all day 2) battery efficient enough to last at least 4 hours between battery swaps or recharges 3) inexpensive enough to be affordable when scaled up to 15 devices/field (x7-15 fields) and also be affordable for African research (with the help of rich world grants) 4) flexible enough to work with different DGPS bases in every field. 5) easy and robust enough to be implemented by anyone. 6) compact enough so the user has hands free to do phenotyping
This system will be able to serve as the backbone for remote sensing tools, potentially through the Android Open Accessory Development Kit and Arduino microcontrollers.
Barcoding & Phenotyping Overview
Our phenotyping system utilizes barcodes to eliminate keying errors while maintaining ease of use. It is also very conducive to high throughput. In the field we use rugged lightweight cordless scanners (Symbol P360 w 4 Mb memory, ~$700). With the cordless scanners we begin each day by scanning five pieces of information: date, scorer(s), location, tool and scanner number. At the end of each day of data collection, we download the data to a computer. These files are then renamed using a script which uses date, scorer(s), location, and scanner number to create a unique name. Using a small application we then upload the contents of each file as a clob into a database to provide yet another backup copy. The contents of the files are parsed and uploaded into a table in the database. When in the lab and able to be connected to a computer, a corded scanner is used. A Symbol LS 2208 scanner can be acquired for < $200. For printing up labels we have had success with the Sato CX400. At times we print >10K labels in a week which would burn out lesser label printers. All possible phenotype values are encoded as barcodes, thereby creating what we refer to as a tool. PDFs of our tools can be downloaded here. For measuring length we create a measuring stick of stacked barcodes generated using our small custom application called Stacked Barcode Creator. For measuring continuous variables such as weight, we create enough barcodes to have one at each value of significant digits. With continuous values such as color or disease resistance that may have a subjective aspect, we typically limit the number of classes or bins to five. This is generally the limit of the number of clear declarative sentences that can be written that a second scorer would have any hope of objectively following. Then if 5 classes is not sufficient, greater granularity is achieved by repeated scoring – even if by the same scorer – and averaging the scores.
