Quality Control Best Practices

Level: Basic

Use Case: As digitisation manager I want to have the quality control in the digitisation process so that I can provide high quality data.

Recommendation:

Establish quality control procedures in all the stages of the digitisation process.

Discussion

Quality control is one of the essential parts in the digitisation process. It will ensure the digitised data is of a high quality level for different usages. In each step of the workflow of the digitisation process, quality checks have to be performed in time to find out the errors and mistakes and alerts for checking and re-digitisation. That will prevent the expansion of the errors to the following steps of the digitisation process and minimise the efforts of corrective actions.

Regarding ETL procedures in the digitisation process, quality control mainly involves two parts, image check and specimen data check. There is still a large amount of manual work involved in the check due to the corresponding work in the workflow that has not been automated. Automating the digitisation workflow and related quality control will reduce human mistakes and improve the work efficiency in the digitisation process.

Implementation

Meise Botanic Garden

MBG implemented a digitisation workflow to digitise the herbarium sheet specimens. This workflow is based on modular designs containing tasks of inhouse and outsourced digitisation, processing, preservation and publishing. In each of the tasks, there are quality concerns that quality control has to be performed. Full details of this workflow can be found in Nieva de la Hidalga et al. (2020).

References

Hardisty A, Saarenmaa H, Casino A, Dillen M, Gödderz K, Groom Q, Hardy H, Koureas D, Nieva de la Hidalga A, Paul DL, Runnel V, Vermeersch X, van Walsum M, Willemse L (2020) Conceptual design blueprint for the DiSSCo digitization infrastructure - DELIVERABLE D8.1. Research Ideas and Outcomes 6: e54280. https://doi.org/10.3897/rio.6.e54280

Nieva de la Hidalga A, Rosin PL, Sun X, Bogaerts A, De Meeter N, De Smedt S, Strack van Schijndel M, Van Wambeke P, Groom Q (2020) Designing an Herbarium Digitisation Workflow with Built-In Image Quality Management. Biodiversity Data Journal 8: e47051. https://doi.org/10.3897/BDJ.8.e47051

Nieva de la Hidalga A, van Walsun M, Rosin, P, Sun X, Wijers A (2019) Quality Management Methodologies for Digitisation Operations, ICEDIG Project Report. https://doi.org/10.5281/zenodo.3469521

Level: Basic (with advanced/state of the art recommendations)

Use Case: As a digitisation manager I want to have the quality control in the digitisation process so that I can provide high quality data.

Recommendation:

Establish quality control procedures for images.

Discussion

Specimen imaging data are one of the key outputs from the digitisation process. It is critical to keep the image quality at a high level. The quality control for the images involves the image acquisition, processing, and storing processes.

In the image acquisition process of mass digitisation, the images are captured and usually transferred to the imaging station immediately. The captured images need to fulfil the following quality control checks:

• Format validation
• File integrity check
• Image size, resolution and metadata verification
• Image colour check
• Image sharpness check

The above quality control measures can be done automatically by the computer applications on the imaging station in real-time. This will find the error images in time for the re-imaging process of the specimens.

After passing the above quality controls, the original images are ready for the imaging processing tasks in the workflow, like image renaming by decoding barcodes in the image and image transformation to other formats. In the image processing process, the quality control measures can be done at the imaging station or on a remote server depending on the workflow as

• Image file name format verification
• Image derivatives (such as JPEG and PNG images) check (similar QC measures that were done in the above image acquisition part)
• Image duplication check

The above quality control measures can be done automatically by the computer applications. Often they can be performed offline, since the imaging tasks are based on the original images and do not need access to the physical specimens.

After image acquisition and processing, different versions of the image are transferred and stored at different storage areas, such as the staging area, image archive, and long-term data preservation. Also the storage of the imaging station and buffer server have to be cleared periodically after the successful image transfer to other data storages by checking

• File integrity

Some of the above complex QC tasks done by computer applications belong to the ADVANCED level.

Moreover, for different types of specimens, there are different objects in the image, such as specimen, labels, colour chart, scale bar, and barcode, as summarised in a table from Hardisty et al. 2020 (see Table 9). It is necessary to make sure that those objects are shown in the image correctly. This work is usually done manually before the digitisation or during the barcoding process. With the development of computer vision and AI techniques, the computer program can achieve relatively high accuracies to detect those objects. However, it requires a large training dataset and computing resources to train the mode and perform the task. This belongs to the STATE-OF-ART level.

Implementation

Meise Botanic Garden

At Meise Botanic Garden, quality control measures are applied to the images in the image acquisition and processing subtasks, and image storing subtasks in the digitisation workflows of herbarium sheet specimens.. Full details can be found in Nieva de la Hidalga et al. (2020)(see Table 3).

References

Hardisty A, Saarenmaa H, Casino A, Dillen M, Gödderz K, Groom Q, Hardy H, Koureas D, Nieva de la Hidalga A, Paul DL, Runnel V, Vermeersch X, van Walsum M, Willemse L (2020) Conceptual design blueprint for the DiSSCo digitization infrastructure - DELIVERABLE D8.1. Research Ideas and Outcomes 6: e54280. https://doi.org/10.3897/rio.6.e54280

Nieva de la Hidalga A, Rosin PL, Sun X, Bogaerts A, De Meeter N, De Smedt S, Strack van Schijndel M, Van Wambeke P, Groom Q (2020) Designing an Herbarium Digitisation Workflow with Built-In Image Quality Management. Biodiversity Data Journal 8: e47051. https://doi.org/10.3897/BDJ.8.e47051

Nieva de la Hidalga A, van Walsun M, Rosin, P, Sun X, Wijers A (2019) Quality Management Methodologies for Digitisation Operations, ICEDIG Project Report. https://doi.org/10.5281/zenodo.3469521

Level: Basic (including an Advanced recommendation)

Use Case: As a digitisation manager I want to have the quality control in the digitisation process so that I can provide high quality data

Recommendation:

Establish quality control procedures for specimen data

Discussion

Specimen data is the most important core part in the digitisation process along with specimen imaging data. In mass digitisation, usually preliminary specimen data with minimum information level are extracted from the specimen during the barcoding process to speed up the digitisation. More extensive transcription can be done later with the specimen image at a collection management system or dedicated transcription portals. To improve the quality of the specimen data in the digitisation process, the quality control measures must be applied in the digitisation process, such as simple data format validation of:

• Date, time and higher locality names

By further utilising the list of controlled vocabularies/terms from the authorised sources to validate:

• Scientific names
• Localities
• Peoples' names

Automated geo-referencing processes can be used to improve the data quality.

Implementation

Finnish Museum of Natural History (Luomus)

At Luomus, the preliminary specimen information is recorded at the barcoding step in the mass herbarium digitisation process with a web-based system. In the system, the list of controlled vocabularies from the authorised sources is used to validate the scientific name, country, and municipalities of the specimen. The special cases of the localities are alerted in the system with highlights to the user. The formats of year, and the links between country and municipalities are validated instantly after the input of the fields. At the mass digitisation of the pinned insect, automated geo-referencing is used to achieve high specimen data quality.

References

Hardisty A, Saarenmaa H, Casino A, Dillen M, Gödderz K, Groom Q, Hardy H, Koureas D, Nieva de la Hidalga A, Paul DL, Runnel V, Vermeersch X, van Walsum M, Willemse L (2020) Conceptual design blueprint for the DiSSCo digitization infrastructure - DELIVERABLE D8.1. Research Ideas and Outcomes 6: e54280. https://doi.org/10.3897/rio.6.e54280