Author：CHANG CHIH-HSIEN,Coastal and Offshore Resources Research Center,Fisheries Research Institute, Council of Agriculture,Executive Yuan
Cage culture has a history of decades. The current marine cage culture farms in Taiwan concentrate in Pingtung and Penghu, mainly producing species like pompano fish, groupers, and cobia (Rachycentron canadum). Compared with land-based aquaculture, marine cage culture has a higher risk and takes a longer production cycle. Because of the properties of the marine water medium, measurement of fish body size is done mostly by either manual sampling or underwater photography. The manual sampling may cause stress, injuries or even casualties of the fish. Underwater photography, on the other hand, has limited by the frequency of SCUBA diving operations. In terms of the amount of fish in the cage, despite the accurate amount known upon stocking, the precise estimation based on visual observation during breeding is hardly possible. The exact amount of the school of fish in the cage is known only upon harvesting, usually with errors in the estimated productivity during the farming process.
We developed the technology for monitoring the length of cage-cultured fish in 2020 and have made use of the technologies developed in the smart agriculture projects in the past few years. The purpose is to help cage culture farmers to exercise more effective production management and scheduling and to monitor biological parameters of the fish like the body size and growth dynamics. With underwater image recognition and AI technologies, we probe into the problem of difficulties of defining the growth dynamics of the cultured fish and set the objectives of measuring the bio parameters of the fish size and quantity. The underwater visual information of the school of fish is converted into digital data to assist in the establishment of standardized aquaculture production, which is more practical and convenient for operators.
As shown in Figure 1, the length monitoring system for cage-cultured fish is portable for boats, and is capable of judging the body size and growth of fish, recording the moving images, and remote monitoring. The system includes the 3D binocular underwater camera, a portable video relay station, and a cage-side frame structure. Its application is adjustable to suit the actual types of the aquaculture environment. For the procedure of operation, the portable video relay station is first set up on the work boat. Then the 3D binocular underwater camera is set up beside the cage by the cage culture farmers when they do their regular jobs of feeding and checking. The images captured by the camera are stored in the portable video relay station, which, upon the return of the work boat, uploads the information to the cloud-based system for storage. There, AI computing technologies are used to analyze and measure the data.
This system has been tested on-site in a private cage culture farm (see Figure 2). As shown in the diagram, the blue frame on the AI screen is for identification of whether the image is that of “a fish”. The fish image identified is selected in the green frame for AI measurement of the length. The data of the body length is further processed with the data of length and weight ratio obtained from references or empirical tests to yield the estimated weight value. The information is finally presented at the site and on the user interface at the monitor center, so that cage culture farmers can monitor the latest growth information of their fish on mobile devices.
The fish body size is a parameter of ultimate importance in aquaculture production. The system now can detect fish size in indoor aquaculture ponds and cage culture farms, and the portable system for boats can be used in changeable marine conditions. The underwater images obtained by this system can further measure the biological parameters of the school of fish, including their length and weight, etc., to gradually replace the current measurement methods of manual sampling and underwater diving photography. This can save the labor cost and minimize errors of manual sampling and stress for the fish. Besides, underwater image analysis will help monitor the weight gain, the group density, the mobility, and diseases of the fish, to serve as basis for the decision on the types of feeds used and the timing of feeding for more efficient aquaculture production management.