Original papersIMAFISH_ML: A fully-automated image analysis software for assessing fish morphometric traits on gilthead seabream (Sparus aurata L.), meagre (Argyrosomus regius) and red porgy (Pagrus pagrus)
Introduction
According to the UN Food and Agriculture Organization (FAO), the global production of food must continue increasing in order to satisfy the increasing demand (APROMAR, 2014). Food from marine origin is an important part of people protein demand; and, currently, more than the half of this food comes from aquaculture. In addition, consumers’ awareness is also increasing, and they demand high quality food. Gilthead seabream (Sparus aurata L.), meagre (Argyrosomus regius) and red porgy (Pagrus pagrus) are marine species well established in consumers’ eating habits. Gilthead seabream is the most important Mediterranean aquaculture marine species: its total production reached 179,924 tons in 2013 (APROMAR, 2014). Meagre and red porgy constitute new aquaculture species and have been recently cultured in the European area, since they are well known and appreciated species among consumers and, additionally, their production systems are well adapted to farming conditions.
There is an increasing need for aquaculture companies’ to improve their competitiveness by optimizing their production and/or by increasing their products quality. The standardization of non-invasive measurement methods and their application on industrial processes minimizes costs, increases the fish measurement rate and maximizes their products added value. This challenge can be partly addressed by investing on scientific knowledge and technological innovation. In that context, biotechnology, together with genetic improvement, plays an especially relevant role. The positive results of considering genetic improvement programs in domestic animal species are remarkable, as annual rates of genetic improvement of 1–3% have been described (López-Fanjul and Toro, 2007). In aquaculture, genetic improvement programs have been successfully implemented in seawater fish, such as salmonids, cyprinids, tilapia, catfish (Gjedrem, 2012). In marine aquaculture, the first genetic breeding program in Spain for gilthead seabream (PROGENSA®) has been developed in collaboration with six Spanish aquaculture companies and four research centers (Afonso et al., 2012, Lee-Montero et al., 2015). In meagre, the first genetic parameters for growth traits have been estimated (Soula et al., 2012a); and, in red porgy, physical tagging systems and multiplex PCRs for the genetic identification have been developed to be introduced in breeding programs (Navarro et al., 2008, Soula et al., 2012b). However, to improve selection efficiency, traits information must be recorded in an objective and precise manner (without personal bias) and at a low cost (Gjedrem, 1997).
Image analysis technologies allow the objective measurement of lineal and dimensional morphological traits, as well as a fast report of reproducible and reliable information. This is useful for genetic improvement programs, which require large number of samples. Another advantage of using image analysis is the availability of images. Collections of digital images can be posted online and serve as easily accessed and sharable archives for researchers and fish farmers (Petrtýl et al., 2014). During the image analysis process, objects are distinguished from background, and numerical information is produced from the captured image. This process includes the following steps: image acquisition, image digitalization, image improvement, image segmentation and measurement operation (Dowlati et al., 2012). Computer vision is the construction of explicit and meaningful descriptions of physical objects from images. A computer vision system generally consists of illumination, a camera, computer hardware and software.
The application of image analysis technology is increasing in the food industry. It replaces the subjective human vision by automatic processes of image analysis for the evaluation of shapes and colors of fruits and vegetables (Du and Sun, 2004, Costa et al., 2011). In aquaculture, there are many potential applications for this technology, which could improve the product quality and the production efficiency (Zion, 2012). An increasing number of studies include image analysis for different purposes in fish farming (Zion, 2012). Several image-analysis-based softwares have been developed to evaluate color and fillet fat content in salmonids and cichlids production (Korel et al., 2001, Stien et al., 2006, Kong et al., 2007, Yağız et al., 2009). Many softwares have been developed in order to identify species (Storbeck and Daan, 2001, Trapani, 2003, White et al., 2006, Zion et al., 2007, Hosseini et al., 2008), to count living fish (Merz and Merz, 2004), to evaluate shape variation (Loy et al., 2000, Blonk et al., 2010, Costa et al., 2015) or to estimate the fish size (Martinez-de Dios et al., 2003, Costa et al., 2006, Balaban et al., 2010a, Balaban et al., 2010b, Torisawa et al., 2011). All these softwares have proven to be successful. However, many of these softwares do not offer the information in a totally automatic manner from image analysis, they require manual establishment of certain specific points. This is a great limitation in large scale studies, such as a genetic selection programs.
In view of the above, the main objective of this study was to develop a fully-automated software for image analysis, in order to determine morphometric traits in three commercially interesting fish species: gilthead seabream, meagre and red porgy; and to validate this software by analyzing 500 fish of each species (total 1500 fish). This would provide fish farmers and researchers with a highly efficient tool to be used within genetic breeding programs or in large scale assessments of fish product. Another additional objective was the development of an adequate and reproducible protocol to capture images.
Section snippets
Image capture protocol
Two photographs from each fish were taken: one lateral image and one dorsal image. Photographs were taken without any outer light influence. To get that, a small dark room with metallic structures covered with black plastic film was used. An Olympus digital camera (FE230/X790, Olympus lens 6.3 to 18.9 mm, f3.1 to 5.9, equivalent to 38 to 114 mm on a 35 mm camera) was fixed to the table. The camera has a 3× optical zoom lens and the lens has a maximum aperture of f3.1 at wide angle. However, the
Results
Manually measured fork lengths and body weights ranged from 12.5 cm to 36.6 cm and from 30.8 g to 1145.9 g in gilthead seabream; from 17.5 cm to 58.4 cm and from 65 g to 2343.6 g in meagre; and from 16.3 cm to 29 cm and from 200.6 g to 697.4 g in red porgy, respectively.
Ninety-eight percent of the captures were successfully processed by the software. Five lateral image captures generated error messages: one in gilthead seabream, two in red porgy and two in meagre. These messages were due to capture
Discussion
The automated visual inspection has many advantages in the food industry: consistency, speed, precision and cost-benefits ratio (Brosnan and Sun, 2004). Accordingly, it has been widely utilized in agriculture and in animal production (Murasawa et al., 2008, Osawa et al., 2008, Rius-Vilarrasa et al., 2009). These advantages are even more relevant to genetic improvement programs, since they require the assessment of a huge number of animals in a short period of time. In addition, within these
Conclusions
The software IMAFISH_ML provides fish farmers with an efficient, fast and automatic tool to evaluate the product for industrial purposes, which shall increase accuracy and minimize costs. Moreover, it allows fish farmers to objectively evaluate morphological and growth traits within a genetic breeding program by providing a high number of fast, easily-measurable and non-invasive traits that could be potentially correlated to other production traits. IMAFISH_ML is available free for further
Acknowledgements
The authors wish to thank everyone who participated in the fish assessment, both taking photographs and manually measuring traits. We also acknowledge the comments of the referees, which have significantly improved the text. This study was supported by INNOTECSS project (RTA2013-00023-C02-00; Spanish Ministry of Economy and Competitiveness).
References (49)
- et al.
Heritability of shape in common sole, Solea solea, estimated from image analysis data
Aquaculture
(2010) - et al.
Improving quality inspection of food products by computer vision – a review
J. Food Eng.
(2004) - et al.
Extracting fish size using dual underwater cameras
Aquacult. Eng.
(2006) - et al.
Are trunk lateral line anomalies and disoriented scale patterns in European seabass (Dicentrarchus labrax) influenced by genetics?
Aquaculture
(2015) - et al.
Application of machine-vision techniques to fish-quality assessment
Trends Anal. Chem.
(2012) - et al.
Recent developments in the applications of image processing techniques for food quality evaluation
Trends Food Sci. Technol.
(2004) Genetic improvement for the development of efficient global aquaculture: a personal opinion review
Aquaculture
(2012)- et al.
Genetic parameters of in-vivo prediction of carcass, head and fillet yields by internal ultrasound and 2D external imagery in large rainbow trout (Oncorhynchus mykiss)
Aquaculture
(2013) - et al.
Comparing geometric morphometrics and outline fitting methods to monitor fish shape variability of Diplodus puntazzo (Teleostea: Sparidae)
Aquac. Eng.
(2000) - et al.
Development of two new microsatellite multiplex PCRs for three sparid species: Gilthead seabream (Sparus auratus L.), red porgy (Pagrus pagrus L.) and redbanded seabream (P. auriga, Valenciennes, 1843) and their application to paternity studies
Aquaculture
(2008)
Estimates of heritabilities and genetic correlations for growth and carcass traits in gilthead seabream (Sparus auratus L.), under industrial conditions
Aquaculture
Estimates of heritabilities and genetic correlations for body composition traits and G×E interactions, in gilthead seabream (Sparus auratus L.)
Aquaculture
Evaluation of Video Image Analysis (VIA) technology to predict meat yield of sheep carcasses on-line under UK abattoir conditions
Meat Sci.
Modelling of fillet fat, fillet weight and fillet yield on live River Catfish (Pangasianodon hypophthalmus)
Aquaculture
Fish species recognition using computer vision and a neural network
Fish. Res.
Automated measurement of species and length of fish by computer vision
Fish. Res.
The use of computer vision technologies in aquaculture – a review
Comput. Elect. Agric.
Real-time underwater sorting of edible fish species
Comput. Elect. Agric.
Sorting fish by computer vision
Comput. Elect. Agric.
Prediction of the weight of Alaskan pollock using image analysis
J. Food Sci.
Using image analysis to predict the weight of Alaskan salmon of different species
J. Food Sci.
Cited by (26)
High-throughput phenotyping by deep learning to include body shape in the breeding program of pacu (Piaractus mesopotamicus)
2023, AquacultureCitation Excerpt :Therefore, the main output of this study was the development of first CVS by deep learning in Pacu, which allowed for easy capture of body measurements in live animals, even in field conditions without controlled background and illumination. In the past, several studies attempted to develop an automated CVS in fish with controlled illumination, and fins were manually removed to improve predictions in some cases (Costa et al., 2013; Jeong et al., 2013; Navarro et al., 2016). Recently, significant advances have been made in Nile tilapia, which was the first CVS created for live fish that could be used in environments without controlled illumination conditions (Fernandes et al., 2020).
Genetic parameter estimations of new traits of morphological quality on gilthead seabream (Sparus aurata) by using IMAFISH_ML software
2021, Aquaculture ReportsCitation Excerpt :In addition, this methodology is also non-invasive and hence can be used in vivo, with minor stress-related effects due to handling (Ruff et al., 1995). Navarro et al. (2016) reported a new method for measuring morphology in three species (gilthead seabream, red banded seabream and meagre), through an automatic image analysis software (IMAFISH_ML) for assessing Non-invasive Technological (NiT) traits, related to fish morphology (mNiT) and carcass (cNiT). This method, however, has not been yet evaluated in terms of additive genetic components and genetic relationships with regular industrial traits such as those related to growth.