Pengaruh Penggunaan Convolutional Neural Network pada Aplikasi Sensor Gambar untuk Deteksi Hewan Liar

Abstract

The collection of sensors that can connect to the internet is known as the internet of things (IoT). Sensors collect data and send the data to a server on the internet for further processing. Sensors can be numeric, image or video sensors. Image and video sensors that use cameras can be used for monitoring applications at home, offices, or other applications such as in the wild. The biggest obstacle in image sensor technology is the energy requirement and the transmission link is quite intensive because image processing in the sensor requires large energy consumption and transmission capacity requires a high bit rate. To reduce the constraints in image sensor technology, an effective application and internet protocol is needed. Local image processing applications can take advantage of soft computing technologies such as machine learning. Meanwhile, protocols can use standard TCP/IP protocols such as user datagram protocol (UDP) and transmission control protocol (TCP). There are many applications that are able to minimize the number of image frames that must be sent by utilizing target detection applications. However, the risk of using this application is the increase in energy consumption of sensor nodes. An example of a camera technology that uses soft computational methods to detect images is the JeVois camera. The JeVois is a miniaturized smart camera, has random access to image data, an easy-to-read mechanism, and has high-speed imaging. Meanwhile, to reduce transmission power, as well as bandwidth requirements, image transmission can be controlled with the detection results of the soft computing method above. This thesis examines the performance of the image sensor using a convolutional neural network (CNN) technique to detect wild animals, and transmits only based on the detection results. If no object is found, then the sensor is set not to transmit an image. Sensor performance is measured by analyzing the energy consumption of the sensor due to energy processing during transmission, the parameter values for packet loss, delay, and jitter are obtained.