Advanced search
Start date
Betweenand

FishTrader: market intelligence platform for profit maximization of fish farmers and the fish industry

Abstract

In fish farming, both producers and meat industry are harmed by the lack of products and technologies to optimize productive and economic resources in the production chain. Among the main difficulties of fish farmers, we can highlight two points: i) lack of control of the growing and fattening process and ii) lack of information when negotiating with the industry. The lack of daily control of growth and supply adjustment of fish feed intake, besides thedamage, can cause eutrophication of water bodies and the appearance of pathogens. For the industrial segment, the lack of standardization of fish that arrives in the industry, information on traceability, quality, volume, planning, logistics and predictability of animal supply, affect this sector. These challenges turn into opportunities in this project, which aims to collect data and create computational models for the development of a market intelligence platform (named FishTrader). The @Tech team has already proven this concept with BeefTrader (Process FAPESP 2015/07855-7), which is an intelligent trading platform that uses daily monitoring sensors (cameras, scales and market information) to maximize the profit of feedlot up to 34%. More recently, in 2017, @Tech developed a proof of concept for poultry industry named PoultryTrader, with support from global food giant BRF, an award-winning proposal from ABDI, from over 800 candidate startups. The FishTrader will evaluate if the physical biometric measurement currently used in fish farming is equivalent to the estimation algorithm based on computer vision and machine learning applied to camera images, which is the hypothesis of the proposal. For this, this project will be developed in two steps: i) the first step will be through the use of user-centered modeling methodologies. Information will be collected from fish farmers and industry in order to codify the trading rules that exist between them (including logistics, allocation and distribution information); ii) the second stage will consist of: ii. a) using computer vision and machine learning an algorithm will be developed capable of extracting measurements from fish images. The species Oreochromis niloticus (tilapia) and Piaractus mesopotamicus (pacu) will be evaluated under controlled conditions in Piracicaba - São Paulo. Amazonian species Colossoma macropomum (tambaqui); Brycon amazonicus (matrinxã) and Arapaima gigas (pirarucu) will be evaluated in the city of Manaus, Amazonas. Step (ii .b): an algorithm will be developed using physical biometrics data, obtained in laboratory under controlled conditions, in the juvenile and adult phases of the proposed species. The developed algorithm will be parameterized and evaluated using image acquisition technologies at three levels of water turbidity (colorless control, moderate and high turbidity), with the following camera types: 2D RGB (Red, Green and Blue), 3D, infrared and multispectral. Therefore, the project for these treatments will have a factorial design, where the localities will be the local control or blocks (AM and SP; n = 2), and within each block will be evaluated the growth phases (n = 2) of the species (n = 5), types of camera systems (n = 4) and water turbidity levels (n = 3). FishTrader innovation will enable commercial and industry fish farmers to maximize the profitability of their activities. This will generate value for these customers on a commercial scale. The @Tech will charge the animals monitored and traded by the platform, generating revenue for all involved and creating new highly qualified jobs in its operation. (AU)