Animal Repellent System for Smart

Farming using AI and Edge

Computing
 OBJECTIVE OF THE PROJECT
The objective of the project is to design, deployment and assessment of an
intelligent animal intruding detection and repelling system for crop
protection based on AI.

To detect and recognize the animals, as well as generate ultrasonic signals

tailored to each species of the animals.
 Problems Identified
• One of the main concerns of today's farmers is protecting crops from
domestic and wild animals attacks.

• There are different traditional approaches to address this problem

which can be lethal (e.g., shooting, trapping) and non-lethal (e.g.,
scarecrow, chemical repellents, organic substances, mesh, or electric
fences).

• Very expensive with high maintenance costs

 Introduction
• Artificial intelligence is the simulation of human intelligence processes by machines,
especially computer systems.

• Using these technologies, computers can be trained to accomplish specific tasks by processing
large amounts of data and recognizing patterns in the data.

• Every industry has a high demand for AI capabilities – including systems that can be used for
automation, learning, legal assistance, risk notification and research.

• AI Computer Vision for detecting animal species, and specific ultrasound emission (i.e.,
different for each species) for repelling them.
 Abstract
• Agriculture automation has been on the rise leveraging.

• One of the main concerns of today's farmers is protecting crops from wild animals’
attacks.

• Develop a system, that combines AI Computer Vision using DCNN for detecting and
recognizing animal species, and specific ultrasound emission (i.e., different for each
species) for repelling them.

• DCNN model to identify the 90 different animals provided in the data set. Finally, train
and test the CNN model and visualize the results by plotting the confusion matrix.

• Alert the farmers by sending SMS.

 LITEREATURE SURVEY
Year Title Author Methods/Overview Merits Demerits
2021 Monitoring the Y A Roopashree; M Deep Learning method Predict Accurately application trained
Movements of Wild Bhoomika; R is used here for animal for two datasets,
Animals and Alert Priyanka; K Nisarga; detection and tiger and elephant.
System using Deep Sagarika Behera classification.
Learning Algorithm
2021 SocialCattle: IoT-based Yunhe Feng; Haoran To monitor cow social Monitor and track Predict cows only
Mastitis Detection and Niu; Fanqi Wang; behaviors to deploy the cow effectively
Control through Social Susan Ivey; Jayne portable GPS devices
Cattle Behavior Sensing Wu; Hairong Qi; Raul on cows to track their
in Smart Farms Almeida; Shigetoshi movement trajectories
Eda; Qing Cao and contacts with each
other.
2021 Farm Vigilance: Smart Bhanu K. N; Sahana an effort is made to performed better in Detect the animals
IoT System for K automate the activities terms of accuracy. only but not
Farmland Monitoring on farmland such as classify the type
and Animal Intrusion irrigation and animal
Detection using Neural intrusion detection
Network using Convolution
Neural Network
 LITEREATURE SURVEY
Year Title Author Methods/Overview Merits Demerits
2021 Internet of Things- Alfin Fernandha build a monitoring activists can Stray Cats
based Cat Detector Pratama; Aisyah Rahma system for the presence monitor the Prediction
System for Kholifah; Birra Lailatul of stray cats using IoT condition of the cat
Monitoring Stray Cats Nafiisa; Muhammad (Internet of Things) in real-time
Fikri Alfaris; technology through images
Moechammad Sarosa
2020 Animal Behavior Henry Roberts; Aviv Proposes a method for higher degree of findings in the
Prediction with Long Segev efficiently converting predictive accuracy study of animal
Short-Term Memory video of animals at any behavior
length into models
capable of making
accurate behavioral
prediction.
2020 IoT based animal L. G. C. Vithakshana; Applying an IoT based Predicted by the to improve
classification system W. G. D. M. Samankula acoustic classification acoustic Nature. accuracy, a large
using convolutional system was designed number of data
neural network using Convolutional will be used.
Neural Networks
(CNN).
 LITEREATURE SURVEY
Year Title Author Methods/Overview Merits Demerits
2019 A Study on Sensor S. Jeevitha; an animal intrusion alert employed to save Not for Agriculture
Based Animal S.Vengatesh Kumar system is designed by human life from
Intrusion Alert employing wireless animal intrusion.
System Using Image sensors for sending an
Processing automatic alert message
Techniques to both the landowner and
forest officials with an
image
2019 Animal Habit Muhammad Abdu Collisions between decrease the Applied Only for
Monitoring System on Kayumbek vehicle and animal still collisions between Road Sahety
the Road side to avoid Rakhimov; Akhadjon do not lose its wildlife-vehicles
animal collisions with Ibragimov; Jun-Dong significance in terms of
Support Vector Lee; Ryum-Duck Oh traffic safety.
Machine Model.
2018 A Self Induced Aswin. V.P. Ram; A. The development of a An early warning identification
Warning System for Sarath Prakash; A. monitoring system for is sent to the forest specifically for
Wild Animal Irshath Ahamed; K. elephant intrusion can officials as well as elephants.
Trespassing Using Anirudh; M. M. help to track the presence the villagers.
Machine Vision Arvindh; N. of animals over forest
System Nithyavathy prone areas
 EXISTING SYSTEM
• There are different existing approaches to address this problem which can be lethal
(e.g., shooting, trapping) and non-lethal (e.g., scarecrow, chemical repellents, organic
substances, mesh, or electric fences).

• Non-chemical control of pocket gophers.

• 22 rimfire rifle or a shotgun can be used to dispatch woodchucks.

• Some motion-activated water sprayers have been developed that spray birds when they
break the motion-detecting

• Permanent Woven-Wire Fencing.

 Proposed System
AI Computer Vision using DCNN for detecting and recognizing animal species, and
specific ultrasound emission (i.e., different for each species) for repelling them.

DCNN model is applied to train the animal dataset to identify the target, and if an animal
is detected, it sends back a message to the Animal Repelling Module including the type
of ultrasound to be generated according to the category of the animal.

RNN is used to detect the animal using bounding box.

 Software specification
Animal Detection
• Python 3.2
• Flask
• TensorFlow, Pandas
 Concept Flow

Animal Detection

Activate Repellent Alert Intrude Animal

Update Animal intrusion

SYSTEM ARCHITECTURE
Test Sample

Animal Dataset
DL Model DL Model
Preprocessing Capture Live Frame

Segmentation Preprocessing

Feature Extraction Segmentation

Classification Feature Extraction

Animal Not Found Matching Animal Found

Animal Type

Repellent Sound Alert Farmer

Repellent
Farmer
 Modules List
1. AIRep Web Control Panel
2. DL Model
2.1. Training & Test Data Annotation
2.2. Preprocessing
2.3. Animal Detection
2.4. Feature Extraction
2.5. Animal Type Classification
2.6. Prediction
3. Repellent Integration
4. Alert Generator
5. Performance Analysis
 DESIGN/WORKING PRINCIPLE/ALGORITHM OF THE
PROPOSED SYSTEM

1. Deep Convolutional Neural Network

 REFERENCES
[1] M. De Clercq, A. Vats, and A. Biel, ``Agriculture 4.0: The future of farming technology,'' in Proc. World
Government Summit, Dubai, UAE, 2018, pp. 11-13.

[2] Y. Liu, X. Ma, L. Shu, G. P. Hancke, and A. M. Abu-Mahfouz, ``From industry 4.0 to agriculture 4.0: Current
status, enabling technologies, and research challenges,'' IEEE Trans. Ind. Informat., vol. 17, no. 6, pp. 43224334, Jun.
2021.

[3] M. S. Farooq, S. Riaz, A. Abid, K. Abid, and M. A. Naeem, ``A survey on the role of IoT in agriculture for the
implementation of smart farming,‘’ IEEE Access, vol. 7, pp. 156237-56271, 2019.

[4] K. Kirkpatrick, ``Technologizing agriculture,'' Commun. ACM, vol. 62, no. 2, pp. 1416, Jan. 2019.

[5] A. Farooq, J. Hu, and X. Jia, ``Analysis of spectral bands and spatial resolutions for weed classification via deep
convolutional neural network,‘’ IEEE Geosci. Remote Sens. Lett., vol. 16, no. 2, pp. 183-187, Feb. 2018.
THANK YOU
QUERIES…?