DESIGN AND IMPLEMENTATION OF A DEEP LEARNING- BASED DEVICE FOR DETECTING FISH FEEDING DESIRE AND TESTING OF THE DEVICE

• Main Author: S D Ervandame Tarigan, Alsandi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75257
• Institution: Institut Teknologi Bandung

This final project book contains the design and implementation of filters for accelerometers and testing of a deep learning-based fish appetite detector. The device is made up of motion sensors and a camera sensor. The motion sensor uses an accelerometer to detect changes in the water surface to determine the fish's appetite. The motion sensor requires a filter to improve the accuracy of determining the level of fish appetite. The filter is created using Matlab to find the filter parameter values that will be implemented in the accelerometer reading programming. The parameter values from Matlab are also used to implement the filter in the deep learning system. The accuracy of the system after applying the filter is higher than the accuracy of the system before the filter was applied. The filter design is done step by step, starting with data acquisition using the accelerometer, determining the pass-band and stop-band frequencies using Fourier transformation, and then implementing the filter in the accelerometer reading programming and deep learning. In general, the created device fulfills the requirements of solving the problem, which is detecting fish appetite, thereby reducing the amount of wasted feed by improving the accuracy of the existing system through the addition of a filter.