Autonomous Forest Surveillance & Safety System Krishna Sai Varma U, Mohit P, Jithender Reddy J, Metilda Florence S Department of Information Technology S.R.M University Chennai, India krishnasaivarma@hotmail.com Dr. Mohan S College of Computer & Information Systems Al Yamamah University Kingdom of Saudi Arabia s.mohan77@gmail.com Abstract—Forest surveillance means monitoring the forest and preserving its wildlife and vegetation. It includes preventing forest fires, deforestation, monitoring endangered species, alerting in emergencies etc. The current day forest surveillance systems are implemented in many ways ranging from Closed- Circuit Television (CCTV) monitoring, camera traps, fire detectors, drones, and many more. All these methods are heteronomous or semi-automatic requiring a constant human attention and also consume a lot of resources making them inefficient and less reliable. Many of these tasks can be easily automated by machines and other technologies. The proposed project is a single autonomous surveillance system, based on object detection technology. The proposed system is capable of monitoring forest fires, intruders, wildlife etc, all at once and alerts the concerned officials immediately and precisely. The proposed system is implemented on a Raspberry Pi with a camera, fire sensors, and multiple other sensors. This system uses a hybrid object detection system which can be trained to detect specific animals, humans or tools. This helps in automating the monitoring of unwanted visitors, dangerous animals, forest fires, endangered species or restricted areas in the forest. The proposed system can not only store the video feed but can also sends the pictures to your email directly along with real-time video monitoring via the internet which allows the users to monitor from anywhere in the world. It also sends instant alerts to your phone via an SMS in emergencies. It is a cheap single module system which is easily expandable and comes with a controlling software which allows us to control the what animals to be detected and their alert levels and also gives us a primary analysis on various things like forest fires, animal population, trespassed areas etc by collecting data from various monitoring modules creating an information system of forest. Keywords—Surveillance System; Hybrid Object Detection; Raspberry Pi; Forest Fires; Animal Detection I. OBJECTIVE To reduce the manpower in a forest by implementing a fully automatic surveillance system using hybrid object detection and various machine learning technologies to monitor forest by detecting animals, capturing pictures & videos and alerting users in case of forest fires, rogue animals and others using email and message and collect data which can be used for Environmental information system. II. PROBLEM STATEMENT The existing forest surveillance systems cannot operate autonomously. They are either are heteronomous or semi- automatic requiring constant human attention, whose tasks can be easily automated by machines and other technologies thus increasing the likelihood of errors. They also consume a lot of resources and require high operation cost. These systems cannot identify specific targets and the alert mechanism is mostly done manually or semi-automatic making them unreliable. III. LITERATURE REVIEW The present-day forest surveillance is implemented in various manual ways like CCTV monitoring, camera traps and automatic ways like fire detecting sensors and drones, etc. but these are not completely autonomous [8]. most widely used methods include. A. Close-Circuit Television(CCTV) The Close-Circuit Television (CCTV) is a surveillance system where many cameras output to one live visual display, where a human monitors them. The problems encountered with this are that it needs a lot of space & cost to operate, it also produces low-resolution videos and mainly, the system's dependency on a human operator who detects some useful activities. It is defective and expensive to detect every event in the monitor buy human thus increasing the error rate. B. Camera Traps A camera trap is a setup with a camera that is activated remotely which is equipped with motion detection systems, like an Infrared sensor or proximity sensors. When a movement is detected by the sensors they will trigger the camera, capturing any footage in front of the camera. This method is used for decades for capturing wild animals when researchers are not present on the field. The problem with these is there are a lot of unnecessary pictures taken and this requires a human to elevate them. About 75% of these can be blank triggers or unrecognizable pictures and a human must sort through these images. In recent times we use image recognition software to sort these images [4][7] but this type of mechanism doesn't provide real-time alerts.