IoT-Based Anti-poaching Technology to Save Wildlife

Abstract

The objective of this paper is to present several modern technologies being developed and used to prevent poaching of endangered wildlife species in different parts of the world. A study has been performed on the various categories of technological components used to develop instruments that tackle poaching. Our paper explores different projects and systems that prevent the loss of lives of wildlife and/or alerting concerned authorities when the animals might be in danger, so that necessary action and precaution can be taken on time and the poacher can be seized. We also try to provide statistical information related to the rate of poaching of certain endangered species around the world. Finally, a new sensor-based system has been proposed taking into consideration all the advantages as well as the limitations encountered in the existing projects.

1 Introduction

Poaching is defined as an illegal activity involving killing or capturing of animals. Animals are usually poached in order to obtain their body parts such as skin, horns or meat which yield high profits when sold or traded in the illegal market in different parts of the world. Poaching has been a prevalent activity across the world for quite some time. The ignorance of poaching has led to endangerment and even extinction of so many species of fauna. Presently, the problem has been taken up seriously and efforts are being made in order to tackle this menacing problem. Along with preventive actions such as the enforcement of laws against poaching, methods involving diplomacy, negative enforcement, demand reduction, etc. [1]. The use of technology is one such weapon against poaching.

With the development of domains such as Networking, Internet of Things and Artificial Intelligence [2], prevention of poaching has become somewhat easier and more successful as compared to efforts made by manpower alone. This paper focuses on some of the systems based on the domains mentioned above. Some of these systems have been deployed in wildlife sanctuaries and wilderness of India and Africa.

2 Characteristics and Classification of Anti-poaching Systems

Designing an anti-poaching system involves taking into consideration some factors that has a significant impact on its performance. Some of these characteristics include:

• Energy efficiency

  The device should consume minimum power since it operates in remote areas.

• Scalable

  The system should provide low-cost expansion and must operate consistently.

• Robust

  Since the system operates in different weather conditions, it must be robust and require less maintenance.

• Simple mechanism

  The technology must be straightforward and the device itself should be compact. This is especially important when the device is attached to the animal’s body.

Existing proposals and implemented systems that are developed can be classified into four major categories:

• Perimeter-based technologies

  They help keep out intruders from reserved/protected parks and stop animals from escaping the area.

• Ground-based technologies

  Camouflaged from animals and poachers, but it is not easy to maintain.

• Aerial technologies

  They are highly sophisticated but have limited capacity in terms of coverage and require high maintenance. They are also quite expensive.

• Wearable (tagging) technologies

  They are quite useful. The limitation is the difficulties that arise during the installation of the system on the animals.

Throughout the conduction of the survey, it has been observed that one of the major components used in any anti-poaching technology is a sensor. A sensor is a device that records any physical parameter and reports it to the concerned component in the system. A variety of different sensors are available, some popular ones used in anti-poaching systems are listed in Table 1.

Table 1 Types of sensors

Sensors, when combined with other components such as actuators establish a powerful and highly useful network. Such a system is called as Internet of Things. IoT can be described as an expansion of the interconnected engineering system and computer technology [3]. The term was coined by Kevin Ashton in 1998. Besides gaining popularity in domains such as city infrastructure improvement, agriculture, home automation, etc. IoT is now being applied for conservation of the ecosystem.

3 Existing Models and Systems

An example of a sensor-based system has been proposed by Tridip Sarma and Vivek Baruah as a design approach. The system has been proposed keeping in mind the objective of providing required medical attention to injured animals in time and save them [4]. As implementation of CCTV cameras or UAVs¹ are impractical in the dense wilderness, an acoustic sensor plays the elemental role in the proposed system. An acoustic sensor measures the acoustic wave property of sound. In this system, the sensor records the input and transmits it to a computer system through a wireless module. The main components of the system include:

• A high sensitivity sound system that detects sound at various distances.

• A Xbee wireless transmitter following Zigbee² protocol.

• A receiver sending the signal from the transmitter to the computer system also following the Zigbee protocol.

• An Analog to Digital Converter (ADC) for conversion of analog signal to a digital one for further processing.

• A computing system equipped with MATLAB³ and pre-fed data.

• An Op-amp to amplify the received sound for maintaining the quality of the sound.

The sensor aims to detect the sounds of guns used by poachers. When the sensor detects a gunshot, it is received by an op-amp that amplifies the audio signal to prevent loss of quality. It then sends it over to a transmitter which in turn transfers it to a receiver. It is then fed to an ADC which converts the analog signal to a digital one for processing. The computer system uses MATLAB to compare this signal with stored data. It performs a comparison to find the matching signal in order to determine the distance at which the gun was used with respect to the location of the sensor.

The system then issues an alert to the forest officers along with the location of the gunshot. In this way, necessary action can be taken on time and the poachers can also be caught. Data pertaining to the sounds of different types of guns used by poachers and the sound of these gunshots at different distances should be studied and recorded so that the accuracy of the comparison is good.

Another system to tackle poaching of tigers has been proposed by Sharanya Krishnamurthy and Dr. Gayathri S. The system is a wireless sensor network (Fig. 1). According to this paper, tigers are poached for their fur and teeth [6]. Around 50 tigers have been poached in India in 2016 alone and such extensive killing of these felines for a prolonged period has led to endangerment of the specie at an alarming level.

Fig. 1

Wireless sensor network

Coming to the working of the system, it is a wireless sensor network containing distributed sensors that record physical parameters in the deployed environment and send the data that is collected to a central station [7]. The central component of the system is a heart-rate sensor. An increase in the heart rate can be from 2 reasons:

• The tiger is looking for prey, i.e., hunting.

• The tiger is trying to be poached or chased by poachers. This implies the animal is under threat.

Decrease in the heart rate also cannot be neglected. It can occur when:

• It is resting/asleep.

• It could be tranquillized—which is a threat.

The sensor system exists as a collar tagged onto the tiger in habitat. It also has a GPS tracker which continuously monitors the location of the tiger. When an abnormality in heart rate arises, inputs from the heart-rate sensor along with its location-time data from the GPS module is processed in the central processing center and sent as an SMS to the registered mobile at the nearest station where officials are located by the GSM.

A point to consider is the sleep patterns of the animal which needs to be considered to determine whether the animal needs help for avoiding false alarms. The location inputs from the GPS also alerts the authorities when the tiger escapes the forest premises and enters nearby human habitats. This would help the forest rangers to safely return the animal back as well as ensure safety of the humans.

A study by Olifants West Nature Reserve states that approximately 1200 rhinoceros in South Africa are being poached every year. At India’s Kaziranga National Park, this number was approximately 27 in and 4 in Namibia. This has majorly impacted the population of rhinos and has bought the specie close to extinction (Fig. 2).

Fig. 2

Annual Rhino Poaching figures [8]

Noseong Park, Edoardo Serra et al. have designed an Anti-Poaching Engine (APE), which has been deployed at Kaziranga, India [9]. It uses a predictive analysis approach where the algorithm is fed with the data of behavior of the rhino and the behavior of the poacher. The reserve was divided into a total of 1014 cells, covering the entire park. A team collected the following features to formulate the behavioral data related to the Rhinos. The data marked the top 3 blocks closest to a food source, building or road. Elevation points were also considered.

A study was also conducted to identify the patterns and behaviors of the poachers. Combining behaviors of both poachers and rhinos, it was observed that a cell c would be attached when:

• The cell has an elevation less than 394 m.

• The cell is close to a water source.

• There are at least two animals a day at a cell at most one hop away from cell c.

• The nearest source of vegetation is less than two cells (i.e., less than 800 m).

Using this knowledge, the number of rhinos in a cell on the next day could be predicted. Regression models were developed to predict these numbers. Predictions were compared with actual validation dataset. The Gaussian process regression produced the best results.

Such an approach to developing anti-poaching system is observed to be quite advantageous. As the algorithm uses a poacher behavior model rather than a rhino behavioral model, it protects 20% more rhinos than using just the latter. Also, the method does not require any significant increase in funding. Hence it can also be deployed in economically weaker countries and regions.

A study conducted revealed that the population of elephants in Tanzania dropped from 355,000 in 1994 to 180,000 in 1999 and lesser in 2011 [10]. The study also showed that Tanzania is losing about 30 elephants everyday to poaching. According to biologists, the loss of elephants has a significant negative effect on the savannah forests of Africa.

Jamali Firmat Banzi presented an idea in their paper with three different action of responses. The proposed infrastructure involves a sensor fusion form of architecture. Sensor fusion is a software that intelligently combines data from several sensors which improves the system performance [11]. It gives a more accurate result by taking into consideration the various inputs. The system consists of the following components:

• Communication Channel

• Central Computer System

• Mobile Biological Sensors (MBSs)

• Cellular smartphone (Fig. 3).

  Fig. 3

  

  Proposed architecture for flow of data in the system

The MBS continuously sends location data of the animal to the Central Computing system via satellite or GSM network connection. The computer is trained with artificial intelligence algorithms (neural networks) which classifies the animal behavior as normal or unusual in nature. If the analysis results in abnormalities, an alert is sent to the concerned officials. The alert contains the location of the animal and a processed image.

Implementing the usage of UAVs has been a popular option for building poaching prevention systems. One such system that has been implemented is known as Systematic POacher detector or SPOT [12]. SPOT employs the abilities of thermal infrared cameras to identify humans and animals for detecting poachers. The system has been tested and is intended to be deployed in several national parks in Africa. SPOT mainly aims to overcome the tedious task of continuous monitoring of video footage from these cameras. We shall see how it works.

Normally, the usage of UAVs for detecting an object of interest (the poachers in this case) encounters the problems detecting small sized objects mainly with movement. SPOT can automatically detect living beings, i.e., humans and animals in near real time. It consists of two processes—Offline training and online detection. During offline training, the video frames are considered as images. Online detection is based on Microsoft’s cloud service called Azure. The images are sent from the local machine to the AzureBasic virtual machine. Once received on the virtual machine, the objects are detected using faster RCNN.⁴ Since AzureBasic is limited to only one local system and virtual machine, it is scaled using Tensorflow Serving. Kubernetes, a fault tolerant load balancer is used to handle data coming from multiple UAVs. Additionally, tools are added for convenient deployment on the Azure Blob storage. This architecture is called AzureAdvanced.

A prominent issue faced is detection of objects at night. Eye-Spy has been developed to improve object detection. It detects objects based on edge detection. SPOT considers Eye-Spy as used by a novice (ESN) as the user does not have to tune in these parameters and can use the system as is. Finally, SPOT it is tested using old videos and a conservation program called AirSheperd.

An article in the ‘Frontiers in Ecology and the Environment’ titled ‘Passive Acoustic Monitoring as a law enforcement tool for Afrotropical Rainforests [2] focuses on developing an evidence-based adaptivity system. Some monitoring programs rely on the evidence of hunting-related activity, for example, spent ammunition cartridges. The effectiveness of relying on such evidence is not yet established [14].

4 Systems Analysis

The above proposed systems provide a great depth in prevention of poaching and protecting precious wildlife. Some of the limitations and improvements of the studied systems have been observed in this survey which will be discussed here.

The real-time poaching detector using acoustic sensor would be a highly effective method to track down hunters due to its short reaction time and the aid it provides in arresting the defaulters. But the system stands ineffective if the poachers do not make use of guns or other sound producing weapons. Many poachers resort to tranquilizing or trapping their target using mechanical traps, which are nearly noiseless. The system’s performance also depends on the quality of the sound sensor used. Gunshots are very large distances can go unnoticed which is a drawback. A solution to this system is to accompany the acoustic sensor with some other sensor device to receive inputs. A visual input would be a good option. More on that will be proposed in the next section.

Systems which rely on network connections such as GSM are infeasible in remote areas where mobile network is unavailable. Moreover, the signal radiations have a negative impact on creatures such as birds and insects living in the forests [15]. Using satellite-based network system becomes an expensive affair and is not suitable for countries with weaker financial statuses. Such systems must employ other networking technologies such as radio channel as the communication media [16]. Another point to keep in mind is to ensure no data is leaked during transmission. Sensitive data related to location of animals should not fall into the hands of poachers.

Anti-poaching systems designed based on behavior of animals and poacher is a very inexpensive and effective method to be adopted. With growth and development in fields of artificial intelligence and data analysis, systems can be made smart to effectively study and target poachers with greater accuracy. A limitation of such system is the unpredictable nature of poachers and animals alike. Also, systems designed for a specific case such as for a Rhino, as seen above cannot be applied to different geographic conditions and habitat. Hence, it limits the usage of the system.

5 Proposed System

The survey over some of the proposed systems has given a sufficient understanding to formulate the overview of an anti-poaching system (Fig. 4). The system will aim to have different levels of security. This enables a more robust system and there are smaller chances of security breach at protected wildlife reserves. The system will consist of surveillance cameras with night vision capabilities at the perimeter of reserves. The night vision cameras will be integrated with infrared cut-off filters so that they don’t suffer any loss of daytime image quality. Installing them at outer boundaries will not cause problems related to maintenance of the cameras. Roads that are surrounding the reserve are gateways to poachers. A good option to monitor roads would be magnetic sensors in the form of ground-based surveillance. They would detect presence of any vehicles which would alert the system. A warning message could be issued so that if the vehicle is not a threat, action need not be taken. Employing sounds sensors within the reserve would be the best way to detect poachers that make use of guns. It would ensure quick and timely action is taken to arrest poachers and save the animal’s life if it is in danger. Biological sensors, in the form of tagged system will continuously monitor the animal’s health. The animals will have to be tranquillized for attaching or repairing the biological sensors. Data is continuously sent to the central computing system where it is analyzed for abnormalities.

Fig. 4

Architecture of proposed system

The central computing system is the heart of this model. Its core is an intelligent algorithm that is continuously receiving and learning from the data it receives. A good choice for the algorithm would be a semi-supervised machine learning algorithm. Such an algorithm would take a small percentage of labeled data, i.e., the historical data such as animal behavior, poacher behavior, etc. The bulk of data would be unlabeled, which is the live data received in real time. Further survey would be required on semi-supervised machine learning algorithms in order to determine the most feasible one for this system. A database consists of facial images of permitted individuals in the reserve area. This data will be stored on a protected platform, such as a private cloud setup for securing data. Upon receiving an image from surveillance camera, the algorithm compares it with the images in the database. If no match is found, it issues an alert including the location and processed image of the unidentified individual. When sound data of gunshots are received, the algorithm calculates the location by comparison with pre-fed data. It immediately issues an alert along with the location. The health status of the animal is also reported. This is determined by the data received by the biological sensor. When data received is abnormal it is also an indication that attention should be given. The message is transmitted to the receiver, i.e., the forest official via radio transmission. The message is transmitted to all the stations irrespective of its distance from the point of attack. Digital data is transmitted over radio channel by modulating it with techniques such as PSK. Another precautious feature that can be implemented in the MBS could be perimeter checking. If the location of the animal is found to be outside permissible reserve boundaries, an alert can be issued to bring the animal back to safety. This also serves to protect nearby human habitats.

6 Discussion

In the proposed system, with multiple sub-systems in the model, it becomes a challenge for poachers to kill or escape. The system ensures both wildlife protection and seizing criminals. The most important attribute is the versatile nature of the system. Some of the components which are not feasible for a particular ecosystem can be excluded. The complexity and level of intelligence of the AI algorithm is a pivotal to the system. As the system mostly makes use of maintainable and inexpensive components, they can be deployed easily.

7 Conclusion

The protection of our wildlife is the need of the hour. Endangerment and extinction of animals is not just a loss but also a cause of drastic irregulates in global environment. As concerned citizens of the world, we must design systems keeping in mind they do not disturb or harm the ecosystem of the wildlife in anyway, which would rather result in the opposite of what needs to be achieved. Systems must require the least human intervention which essentially means, they must be low maintenance. With robust systems, we can eliminate poaching activities.

8 Future Scope

This paper has given us the insight and an introductory understanding into the technological developments that are being made toward protection of wildlife. The working and the evident drawbacks of some of these systems have also been mentioned when talking about each system. In the future course of wildlife protection, new and more useful adaptations of domains such as IoT, blockchain and data analysis will definitely help enhance the systems. AI algorithms that are currently being used can be improved and with more training, the output would be more accurate. Blockchain has been started being adopted more extensively due to its property of easy scalability and bringing various geographical species under a single surveillance system. The need and development of the system is primarily dependent on the requirement of the region, but certainly is necessary.