Automation of plant assessment and visual classification of plant test elements

Currently, the collection and recording of data through physical inspections of plots or test or cultivation areas require a high level of personnel and time effort. Fieldwork is also associated with physical exertion and is further complicated by weather conditions (sun exposure, wind, rain, fog, etc.), chemical exposure, and possible allergies. Standardizing assessments is challenging because observations and evaluations are carried out by trained personnel, but subjective biases cannot be completely ruled out. Furthermore, conducting assessments is very resource-intensive, as they must be carried out during the growing season (depending on the field, 3-4 times a week). This requires corresponding training times, which require additional personnel and time resources. Previously used devices include a measuring device for recording plant height and the skills of authorized persons who record and evaluate the data. In the field of plant health, rapid detection of changes in the plant population is also necessary. A standardized and automated collection of various values for plant health (photosynthesis index, color shifts, drought stress, etc.) can enable rapid detection of plant diseases or the occurrence of pests in the field, allowing timely control measures to be taken.

Ideally, the recording of the assessment or collection of plant health values is automated, with the responsible person observing the machine's recordings. Ideally, this represents a significant physical relief (e.g., in measuring growth height, counting the number of plants per m², etc.). Plausibility tests should be possible on-site to assess the quality of the recordings directly on-site.

The minimum requirements (must criteria) for this technology are as follows:

• Since the collection of plant data takes place in an outdoor area, the use of the hardware is characterized by constant weather resistance, mobility, and offline usability.

• Recognition of the crop by a camera system and counting or estimating the population on a defined area (plant counting).

• Determination of plant population density. Recording the growth height of the measured crop (this varies greatly depending on the crop - from a few cm to about 3m height for corn).

• Recording the flowering time of a crop. At least five to ten plant characteristics per crop can be recorded as "minimum criteria" (this is necessary to determine, for example, the variety or non-variety-identical plants).
  color,
  shape (e.g., shape of the ear pyramidal),
  posture (e.g., posture of the ear upright or hanging),
  number of side branches,

• the possibility of releasing beneficial insects in the field,

• localization of individual plants or recording using GPS data (e.g., direct localization during sowing) and the possibility of processing GPS data.

• Recognition and differentiation of plant species (crops from each other, cross-pollinators, and weeds, etc.)

• Recognition and differentiation of varieties of a crop species (e.g., crop species wheat, variety Josef, variety Philipp).

• Differentiation between crop stands (e.g., recognizing distances to other crop species).

• Conducting an assessment according to predefined routes in the plant population or recording a specific area or several sub-areas within a field.

• Visual remote evaluation of plant health (e.g., infrared spectra, photosynthesis index NDVI). Recording changes in the plant population (drought stress, water stress, feeding damage, etc.).

• Extended requirements (= should criteria) include: Rapid population assessment even after damage, e.g., damage from fungi/insects/feeding damage, recording soil and leaf conditions after drought or heavy rain events.
  Note: Weather changes the population and appearance (e.g., fog, rain, wind, clouds, and sunlight due to shadow formation).

• Recognition of so-called "deviants" (non-variety-identical plants) in a later repeated assessment (identification of a data set in the real field).

Up to 30 characteristics per crop are available, which are only visible at specific times depending on the plant's development stage.
Note: Over 60 crop species can be identified through AGES.

The aim is to achieve as little intervention in the population or cultivated crops as possible (mostly a "drive-through" between the plants is not possible).

Any legal requirements must be considered.

The following have been defined as non-goals:

Limit value check is carried out in AGES systems and is therefore a non-goal in the sought system (= e.g., checking variety purity through certain characteristics).

No manual sampling is required by the system.

Furthermore, no manual maintenance measures are required by the system (such as weed management).

The operation of the new system should be autonomous by AGES and not require ongoing support from suppliers.

AGES is looking for a tool that can carry out data collection in a plant assessment and subsequent data analysis as resource-efficiently and minimally invasively as possible.

In addition to automating the assessment, the physical workload should also be reduced by having the technology perform certain repetitive tasks independently, with humans always having control.

Efficient data collection (recognition, counting, recording of plant characteristics, and damage detection) are minimum requirements. Since the collection of plant data takes place in an outdoor area, the use of the hardware is characterized by constant weather resistance, mobility, and offline usability.

Extended functions: The system should be able to recognize plant species, identify deviations from the standard, and carry out inspection specifications according to certain standards.

In any case, a comprehensive presentation of the tool and its necessary prerequisites is desired so that these can be integrated into AGES's ongoing IT operations. Information on references or experiences from previous projects in image capture and interpretation is also of interest to the client.

Furthermore, the client is interested in whether there are possibilities for a proof of concept (PoC). The documents are completed with a concept that describes the service conditions, maintenance, and support in case of failures.