Automatic fish scale analysis: age determination, annuli and circuli detection, length and weight back-calculation of coregonid scales

Vogelmann, Christian; Schock, Christian; Feilhauer, Marius; Stibor, Herwig; Schubert, Michael (2025): Automatic fish scale analysis: age determination, annuli and circuli detection, length and weight back-calculation of coregonid scales. Ecological Informatics, 92: 103517. ISSN 15749541

Abstract

Age determination and growth measurment through scale analysis offer valuable insights into fish population structure and growth history, providing indications of abiotic, anthropogenic, or evolutionary influences. Traditional scale analysis is time-consuming and reliant on expert personnel. This study introduces an automated or semi-automated approach for analyzing coregonid scales, highlighting the benefits of digitization. Compared to manual stereomicroscope examination, software-assisted analysis enables reproducible and independent evaluation. Biometric features such as scale area, radius, circuli count, and spacing can be measured faster. Automatic detection of the scale center and standardized determination of the longest radius allow for efficient back-calculation of fish length, weight, and age. Two independent datasets were used to assess the performance of the automated system. Results from paired validation analyses (n = 84 scales) demonstrate high agreement between automated and manual measurements for scale area (r = 0.97, r2 = 0.95) and circuli count (r = 0.78, r2 = 0.60). In a larger field application dataset (n = 1095 scales from 124 fish), the automated algorithm identified the scale center without manual correction in 77.9 % of cases. For the remaining 22.1 %, the mean correction was only 0.15 mm (median: 0 mm), and just 7.6 % required a correction greater than 0.5 mm. Deviations between fully automated and manually verified results for circuli count, radius length, circuli spacing, and age determination were consistently low, supporting robust performance under real-world monitoring conditions. All quantitative metrics are reported in detail in the study. Back-calculated length and weight across age classes were consistent between methods.

Manual processing of 84 scales required approximately 16 working hours, whereas the automated workflow—including scanning and analysis—was completed in under three hours. For the complete field dataset of 1095 scales, automated analysis required only a few minutes after scanning, while full manual verification was completed in less than two hours. These results demonstrate a substantial gain in efficiency, enabling highly standardized, reproducible, and large-scale application of scale analysis in ecological monitoring and fisheries management.

• Dokumententyp: Artikel (LMU)
• Organisationseinheit (Fakultäten): 19 Biologie
• DFG-Fachsystematik der Wissenschaftsbereiche: Naturwissenschaften
• DFG: Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) - 491502892