Vision and sensor-based modeling of intelligent on-demand fish feeding machine
Currently, handfeeding is the most accurate method for fish feed distribution, as the responses of the fish can be directly observed. However, it is labor-intensive, costly, and needs skilled laborers. The time-based feeding is inaccurate as it does not reflect the actual feeding responses. In its e...
Saved in:
Main Author: | |
---|---|
Format: | text |
Language: | English |
Published: |
Animo Repository
2022
|
Subjects: | |
Online Access: | https://animorepository.dlsu.edu.ph/etdd_ece/3 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1003&context=etdd_ece |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | De La Salle University |
Language: | English |
Summary: | Currently, handfeeding is the most accurate method for fish feed distribution, as the responses of the fish can be directly observed. However, it is labor-intensive, costly, and needs skilled laborers. The time-based feeding is inaccurate as it does not reflect the actual feeding responses. In its essence, the technological advances for intelligent precision fish feeding still need to be developed. For instance, the on-demand fish feeders using sensors are prone to false hunger detection. Furthermore, the existing behavior-based feeding method using computer vision (computer vision) has several challenges, including real-time operation, costly hardware requirements, and low accuracy in turbid water conditions.
In this study, an on-demand intelligent fish-feeding machine was developed to optimize autonomous operation for feeding management systems in recirculating aquaculture systems. A fusion of emerging technologies was utilized to minimize the drawbacks of each technology, namely computational intelligence, computer vision, sensors, and the Internet of Things (IoT). Specifically, a fuzzy inference system was used for the decision support system (DSS) to autonomously feed the fish using the accelerometer sensor as the direct triggering mechanism to activate the feed dispenser. In contrast, fish feeding behavior and excess feeds were quantified using computer vision and deep learning, which served as the second feedback control to eliminate the feeding decision errors from the sensor. Other input parameters were considered in the DSS, such as the total dispensed feeds and the time duration of the latest dispensed feeds. The measures of the latter two input parameters were made possible with Edge computing and IoT. Moreover, an IoT-based water quality monitoring system was also developed to ensure that the water is within the ideal condition, specifically for tilapia.
With the developed system, significant contributions are noted in the context of optimizing fish feeding management: (1) implementation of DSS for monitoring and control in a remote and real-time manner with an IoT visualization dashboard, (2) intelligent and autonomous operation of the fish-feeding machine, and (3) technical advances and improvement of fish feeding behavior recognition. In the context of the gathered data from the monitoring system, further insights can be leveraged for prediction, diagnosis, and anomaly detection relating to feeding management. |
---|