Measuring the total biomass of aquaculture production is a critical aspect of aquaculture management as it enables farmers to monitor the growth and development of their aquatic organisms, adjust feeding regimes for optimal growth, and plan harvest schedules (Chen et al., 2022). Although a common practice, manual assessment of biomass estimation at aquaculture farms is time-consuming and usually relies on weighing a sample percentage of animals which may be sensitive to physical disturbance (Khanjani, 2020). The present case study investigates a practical deployment of image-based computer vision systems to support urchin biomass estimation at an Integrated Multi Trophic Aquaculture (IMTA) farm in South Africa. It also discusses the main challenges, requirements and possible limitations to best support aquaculture applications.
The urchin Tripneustes Gratilla is not native to South Africa but is a commercially valuable species for the far Eastern export market. This is one of the new fast-growing, high-value species investigated at the ASTRAL South Africa IMTA laboratory at Viking Aquaculture’s Buffeljags aquaculture farm on the south coast of the Western Cape. The urchins are grown in open baskets placed in large seawater tanks (Figure 1) with active aeration and water circulation to simulate the dynamic inter-tidal conditions in which urchins occur naturally. The prototype developed estimates the diameter of individual urchins in the images using computer vision. The advantages of a computer vision system for urchin diameter estimation include standardization of the process, reduction of skilled operators, and reduced stress on the animal compared to manual measurement procedures. It is a non-invasive tool aligned with animal welfare principles for successful harvests and profitability. The laboratory aims to estimate urchin diameter with up to 10% relative error.
Towards the development of automated measurements of sea urchin size using computer vision sensing, images of the growing urchins are captured monthly. For the development and calibration of the computer vision algorithm, the urchins are photographed underwater in a rigid basket designed to reduce background unevenness in the mesh, therefore providing a consistent reference for calculating scale. Each photographed urchin is measured for its corresponding diameter and mass. In the development phase, measurements for training and validation require that the urchins are removed from the water to measure their diameter using calipers and are placed on a small scale to be weighed. The removal of the urchins from the water is performed only in the development phase of the algorithm, as the computer vision system intends to avoid physical disturbance of the animals in an operational setting. The computer vision algorithms were developed using the Python programming language. Figure 2 illustrates the image processing block diagram.
The image processing step allows the animal diameter to be estimated based on the scale information provided by the crop basket screens/mesh. The HSV color space represents the images. A boundary operation delimits the pixels representing an urchin and allows the removal of the image background. Blur and morphological operators (erosion and dilation) remove noise from the image and improve the representation of the animal. Next the HoughCircles algorithm detects the urchin outlines. Then another step of the algorithm is performed to calculate the circumference of the urchin’s body without the spines. Figure 3 shows the step-by-step of the algorithm.
Results and Discussions
Table 1 presents the results obtained in the current version of the algorithm. The results were divided into two groups, one group comprises the images that do not have ulva over the urchins, and the other group has ulva over the urchins. Ulva is a type of sea lettuce used as food for the urchins and is therefore frequently present in their environment, often sticking to the urchin’s spines. The dataset of the current version contains 33 images. The Ulva make processing the images more difficult, resulting in lower accuracy in the diameter estimation.
So far, improvements are being developed for estimating the diameter of the urchins; after obtaining a better accuracy in this estimate, the next step is to improve the algorithm for estimating biomass, which will have as input parameter the calculated diameter. The technology supports urchin biomass estimation avoiding time-consuming and manual processes. It may also contribute to animal welfare through a non-invasive image-based system.
This work is part of the ASTRAL (All Atlantic Ocean Sustainable, Profitable and Resilient Aquaculture) project. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement Nº 863034.
CHEN, Fudi et al. Intelligent feeding technique based on predicting shrimp growth in recirculating aquaculture system. Aquaculture Research, v. 53, n. 12, p. 4401-4413, 2022.
KHANJANI, Mohammad Hossein; ALIZADEH, Morteza; SHARIFINIA, Moslem. Rearing of the Pacific white shrimp, Litopenaeus vannamei in a biofloc system: The effects of different food sources and salinity levels. Aquaculture nutrition, v. 26, n. 2, p. 328-337, 2020.