Goan students build automatic fish sorting system

With the aim of helping locals and small scale industries to eliminate manpower, save time, and thereby make the segregation of fish more efficient, four students from Agnel Institute of Technology and Design, Assagao worked on developing a fish sorting system which uses image processing. The group comprised of Sahil Deosekar, Gauri Halankar, Amey Patil, and Devashree Samant. The team was mentored by professor, Gauri Gaunekar and professor, Mrunal Sawant.

“On an average fish consumption in Goa is 70 to 90 tonnes a day. Being the staple food of Goans, the demand for fish in the local market has increased and so has the cost,” says Deosekar.

However, according to Halankar, several tasks are involved between the catch of fish to selling it. “One of them is sorting the fish on the basis of different sizes. Since a standard type of net is used, fish of different sizes are caught at a time,” she says.

Segregation of fish based on sizes is a must as the cost of the fish depends upon its size; larger the size more expensive is the fish as compared to smaller ones of the same species. Generally, fish is manually sorted by size using 10 to 20 labourers.

The project built by the team however automatically segregates fish into different sizes. First, a camera captures the fish image. The processor performs all the necessary analysis of the image and calculates the length. Thus, depending on the length, the segregation is done in the sorting part. “The manual sorting cannot be replicated by any machine entirely. The task of sorting requires expert observation skill to slot fish into different categories that only a human eye can observe with accurate precision,” says Samant.

The team took around seven months to complete the project prototype while the research and study about the project took them another two months. “We went to different people like the locals in our village as well as the fish sellers in Mapusa market. We also visited the Betim fish jetty,” informs Halankar. This task was mainly carried out by Samant. Halankar’s role was doing all the software study, software implementation, writing the report along with PowerPoint presentations. Deosekar and Patil largely worked on building up the prototype and studying the hardware components. Deosekar also helped in software implementation as and when required.

Halankar says that their project will definitely help the local fishermen of Goa as commercial fish sorting is labour intensive job; their system would increase the efficiency and reduce the labour effort. The team is now planning to work a little more on their project and introduce the concepts of machine learning in order to make it more impactful and useful for the fishery industry. Once they achieve their goal they will look into how to go about marketing it.

A brief description of the model as follows:

– The prototype used for image processing for the task of fish sorting had a Raspberry processing unit with its operating system.

– The presence of the fish is detected by a proximity sensor. As the fish approaches the desired position on the conveyor belt the sensor triggers the camera to capture the fish image. The processor then calculates the length of the fish; compares it with a predefined database and sends the results to the control system.

– After implementation, the conclusion was that the system was able to achieve an accuracy of almost 90 per cent based on the given testing data, and the total time of execution of one fish was five seconds.

RAMANDEEP KAUR| NT