Description:Title of The Invention
AI AND IOT BASED ANIMAL RECOGNITION, INTRUSION DETECTION AND REPELLENT SYSTEM
Field of the Invention
This invention relates to AI and IOT based animal recognition, intrusion detection and repellent system.
Background of the Invention
KR20160080448A says that present invention relates to a system for protecting crops from wild animals. More specifically, the system for protecting crops from wild animals comprising an unmanned air vehicle eradicating wild animals trespassing on a farm; a sensor unit detecting that the wild animal trespasses on the farm; and a center server receiving a trespassing signal of the wild animal from the sensor unit and transmitting information to the unmanned air vehicle. The system for protecting crops from wild animals is a system of providing data to prepare for the wild animal by not only eradicating the wild animals but also blocking approach possibility and making statistics of the trespassing wild animals. The present invention is to provide the system for protecting crops from wild animals, which is free from restraint on a space and is performed at low costs. The system for protecting crops from wild animals also takes direct measures in accordance with the position of the wild animals. Also, the system for protecting crops from wild animals has double steps for eradicating the wild animals and includes a step of making the statistics on each kind of the trespassing wild animals.
Research Gap: Fails to distinguish human and animals, Costly.
JP2009178140A says that to provide an invasive animal-threatening system capable of detecting at low cost and in high accuracy the invasion and the direction of the invasion of harmful birds/animals with a small units of detectors even under such conditions as to be unable to specify the invading direction and/or position of harmful birds/animals for a specific range wanting to block their invasion, and also preventing harmful birds/animals from being accustomed to threatening by giving a threatening stimulation toward the direction of them detected.

SOLUTION: The invasive animal-threatening system has the following construction and mechanism: A plurality of infrared detectors functioning to detect infrared rays emitted from harmful birds/animals and, in response thereto, output detection signals are installed and arranged on the circumference of a base while partly superimposing the detection views of the plurality of infrared detectors. Infrared rays are detected by which detection view of the detection views formed in the periphery of the base is determined by the arranged infrared ray detectors and thereby the direction where the harmful birds/animals are present is determined. Further, threatening stimulation(s) such as light, sound and (or) odor is (are) emitted toward the direction where the harmful birds/animals are present.
Research Gap: Fails to distinguish human and animals, hence will play alarm even if a human pass by.
US20180125058Al says that invention relates to a multifunctional animal repeller including a central processing unit, and a sound output circuit, a light output circuit, an ultrasonic output circuit, a red light blinking output circuit, a various animal deterrence sound programmer, an ultrasonic frequency selection controller, a multifunctional light selector, an animal repelling mode controller, an operating time selection button, an infrared induction distance adjuster, a photosensitive detector, a camera, a pre-amplification circuit, a mass memory, a WIFI module, a camera effect monitor and a multi-power supply system which are all connected with the central processing unit. The multifunction animal repeller can repel different animals through physical phenomena such as sounds, ultrasonic waves and flashing light based on a humane animal repelling design and can protect animals against injuries, and thus the natural environment is protected.
Research Gap: Fails to distinguish human and animals.
JP2006158372A says that provide a bird- and beasts-repelling method and a device which do not cause reduction/loss of efficacy astonished/feared by keeping birds and beasts so as not to enable to observe/learn/understand the expression conditions of the repelling method.

SOLUTION: In the repelling method, repelling element is not expressed until birds and beasts approach prey in order to block observation/learning of intimidation element expression conditions by birds and beasts and always encounter expression of intimidation element in unlearn/unexperienced conditions and 2-5 m approach of birds and beasts to the prey is sensed by infrared ray emitted by birds and beasts and birds and beasts are made to astonish and fear and repelled by blink of sudden intense light and emission of intimidation sound and expression of repelling elements is stopped after repelling and observation/leaning chance is not given to birds and beasts. In the method of repelling crows, crows carrying out watching/observation at a higher place near the prey which is ordinary aspect for attacking prey are repelled by solar reflection irradiation by a hand mirror or laser pointer irradiation. As a result, serious attack of crows in the following stage is prevented.
Research Gap: Fails to distinguish human and animals.
AU2020103507A4 says that a system and method for crop protection is disclosed. The system comprises an Outer 5 terminal (1) and a Server (2). The outer terminal incudes a light emitting unit (11), a light receiving unit (12), a microcomputer (13), a communication unit (14); and is coupled with a vibration detection (15) sensor installed on the fence that detects the presence of wild animals or birds. This sensed data is communicated to a server (2) configured with the outer terminal. The server includes a control unit (22), a 10-monitoring unit (23), an alarm generating unit (24) and a display unit (25). When the detection signal is more than a set number, an internal monitoring operation signal is generated by the server that is input to the control unit (22) which then determines whether to generate an alarm signal, and to display and process and store the captured image in response to the monitoring operation signal. LZII2 15 Figure 1. The overall configuration of the Outer terminal and server as disclosed by the present invention.
Research Gap: Fails to distinguish human and animals.
None of the prior art indicate above either alone or in combination with one another disclose what the present invention has disclosed. This invention relates to AI and IOT based animal recognition, intrusion detection and repellent system.
SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
Livestock and wild animals can cause significant financial losses for farmers by damaging crops, and this problem is compounded by the fact that many of these animals are attracted to farmland due to the abundance of food, shelter, and water that it provides. The issue of animals destroying farms is especially prevalent in rural areas where farms are often located near wildlife habitats. This not only affects the financial stability of the farmer, but also has a wider impact on the local economy. For instance, when crops are damaged, there may be a shortage of food, which can lead to higher prices for consumers and a decrease in profit margins for farmers. Animals can also pose a threat to the health and safety of farm workers. Cows and horses can become aggressive and charge at people, while deer and squirrels can carry diseases that can be transmitted to humans. As such, farmers are often forced to invest in measures such as fencing and scare tactics to keep animals away from their crops, which can be expensive and ineffective. The problem of animals destroying farms is complex and multifaceted, and it requires a proactive and integrated approach to address it effectively. While there are measures that farmers can take to minimize the damage caused by wildlife, such as using deterrents, planting crops in raised beds, and installing electric fences, these are often not enough to fully protect the crops and the farm. Therefore, the text proposes a computer vision-based approach that can help farmers to monitor wildlife behavior and reduce wildlife damage to crops in a non-lethal and sustainable manner. This approach can provide farmers with targeted and effective measures to deter animals from entering their land, while also preserving the delicate balance between agriculture and wildlife conservation.

BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
Fig A. Outline of the model
Fig B. Working of PIR Motion Sensor
Fig C. Working of Animal recognition model
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
Agriculture is the backbone of many economies, providing food, fiber, and livelihoods for millions of people. However, animals destroying farms is a problem that has been affecting farmers for centuries. Livestock such as cows, horses, and sheep, as well as wild animals such as deer, squirrels, and rabbits, can wreak havoc on crops and cause significant financial losses for farmers. The problem is compounded by the fact that many of these animals are attracted to farmland due to the abundance of food, shelter, and water that it provides. This leads to an endless cycle of damage, repair, and re-damage that can be both costly and time-consuming for farmers. The issue of animals destroying farms is especially prevalent in rural areas where farms are often located near wildlife habitats. In such cases, farmers must contend with both domestic and wild animals that can easily wander onto their land and cause damage. This not only affects the financial stability of the farmer, but also has a wider impact on the local economy. For instance, when crops are damaged, there may be a shortage of food, which can lead to higher prices for consumers and a decrease in profit margins for farmers. In addition to damaging crops, animals can also pose a threat to the health and safety of farm workers. For example, cows and horses can become aggressive and charge at people, while deer and squirrels can carry diseases that can be transmitted to humans. As such, farmers are often forced to invest in measures such as fencing and scare tactics to keep animals away from their crops, which can be expensive and ineffective. Despite the challenges posed by animals destroying farms, there are several measures that farmers can take to minimize the damage caused. These include using deterrents such as noise makers and flashing lights, planting crops in raised beds, and installing electric fences. However, the most effective solution is to adopt a proactive and integrated approach that combines physical barriers, behavioral modifications, and sustainable land-use practices. In conclusion, animals destroying farms is a complex issue that has a far-reaching impact on both farmers and the wider community. While it is a challenge that has been around for centuries, there are steps that can be taken to minimize the damage caused, and to ensure that farmers can continue to produce food in a sustainable manner. To address this issue, a computer vision-based approach can be used to monitor wildlife behavior and reduce wildlife damage to crops in a non-lethal and sustainable manner.
Best Method of Working
A system comprising motion sensors and cameras will be positioned in strategic locations to create a mannequin. When an animal triggers the motion sensor, the camera will activate and use an AI computer vision algorithm to identify the species. Once identified, the system will emit ultrasound emissions that are unique to each species to repel them. This intelligent sensing agriculture system is implemented using embedded edge AI and evaluated for its effectiveness in detecting and identifying various animal species. The system uses Convolutional Neural Networks (CNNs) to conduct the animal recognition process.
The ultrasound won’t affect humans because animals have much greater sound sensitivity thresholds than do humans. They can hear noises with lower frequencies than the human ear can. For instance, whereas humans have an auditory range of 64Hz to 23 KHz, animal such as goats, dogs, cats and elephants have audible ranges of 78Hz to 37 KHz, 67 Hz to 45 KHz, and 45 Hz to 64 KHz and 14Hz to 12KHz respectively. Animals become upset when exposed to ultrasounds that are produced within the critical detectable range, which drives them to flee the sound source. However, even if the frequency band is too high for the human ear, these ultrasounds do not cause any issues. Human eardrums cannot vibrate under ultrasound frequency because they have a much lower specific resonance frequency than animal eardrums.
PIR Motion Sensor: PIR motion sensors consist of pyroelectric sensors that operate based on the principle of Pyroelectricity. Pyroelectricity refers to the property of certain materials generating voltage when exposed to temperature changes, such as heating or cooling. Every entity, whether living or non-living, emits a level of radiation, and the amount of radiation emitted is directly proportional to the entity's temperature. For example, human beings emit radiation at around 12 microns.
The PIR motion sensor consists of two slots made of a material sensitive to Infrared (IR) radiation, along with a lens. Both slots can detect motion within their sensitivity range. In their idle state, when exposed to ambient conditions, both slots detect an equal amount of IR radiation. However, when an object such as a human or an animal approaches the sensor, it intercepts one half of the PIR sensor, causing a positive differential change between the two halves. Conversely, when the object leaves the sensing area, a negative differential change occurs. These positive and negative changes in differential trigger an output signal. This output signal is typically a pulse of approximately 3V in magnitude. In addition to the sensors, the lens on the PIR sensor plays a crucial role in its effective functioning. The lens surrounding the sensors is a Fresnel lens. The Fresnel lens design reduces the amount of material required compared to a conventional lens by dividing the lens into a set of concentric annular sections.
Once the PIR motion sensor detects motion, it triggers the activation of the Raspberry Pi's camera to capture an image. After capturing the image, the camera will record video for a duration of 5 seconds. The frames from this video will be compared with a database using our suggested model (Explained Below). If a matching animal is identified, the system will initiate the release of ultrasound waves as a deterrent mechanism.
Animal Recognition Model: The proposed Convolutional Neural Network (CNN) architecture consists of 1024 pixels, forming a 32x32 image. The CNN is divided into eight blocks, each serving a specific purpose in the network's operation.
A) The input data comprises animal faces from the dataset. To improve computation time, each animal face is resized to 32x32 pixels. The input database is expanded by scaling, rotating, and shifting the data to enhance experimental results.
B) The second block represents a 2D CNN layer with 16 feature maps using a 3x3 kernel dimension. L2 regularization is employed due to the limited dataset, and the Rectifier Linear Unit (ReLU) serves as the activation function.
C) In this layer, a 2x2 kernel is utilized, followed by dropout with a probability of 0.25. This step helps prevent overfitting of the neural network.
D) The second 2D CNN layer employs the same parameters as the first one but with an increased number of feature maps, now totaling 32.
E) A MaxPooling layer is applied, followed by dropout with the same probability.
F) A standard dense layer with 256 neurons is used, and ReLU is applied as the activation function. L2 regularization is implemented for better weight control.
G) Dropout is set to 0.25.
H) The output dense layer consists of multiple classes, softmax activation function is utilized.
In the proposed CNN architecture, pooling operations are applied separately to each feature map. The network repeats this process through successive layers, enabling the system to effectively recognize objects. For instance, the CNN may learn to detect edges in the first layer, use the edges to identify simple shapes in the second layer, and further utilize these shapes to recognize higher-level features like facial shapes in subsequent layers. In the CNN architecture, the neurons in each layer are organized in a 3D structure, allowing for the transformation of a 3D input into a 3D output. Specifically, when dealing with image inputs, the first layer (input layer) represents the images as 3D inputs, where the dimensions correspond to the image's height, width, and color channels. The neurons in the first convolutional layer establish connections with specific regions of these input images and perform transformations, resulting in a 3D output. As the information propagates through the network, the hidden units (neurons) in each layer learn nonlinear combinations of the original inputs, thereby extracting relevant features. These learned features, also referred to as activations, serve as the inputs for the subsequent layer. This process continues, with the learned features from one layer becoming the inputs for the next layer, until reaching the final layer of the network. The learned features obtained throughout the network ultimately function as inputs for the classifier or regression function, which make the final predictions or estimations.
ADVANTAGES OF THE INVENTION
1. Non-lethal and Sustainable Solution: The use of computer vision allows farmers to mitigate wildlife damage to crops without resorting to lethal measures. It promotes a more humane approach to wildlife management while ensuring the long-term sustainability of ecosystems.
2. Improved Resource Allocation: With computer vision technology, farmers can allocate resources more efficiently. By accurately identifying the species responsible for crop damage, farmers can focus their efforts on addressing specific threats, rather than implementing generic deterrent measures. This leads to cost savings and reduces unnecessary disturbance to non-problematic wildlife.
3. Scalability and Flexibility: Computer vision-based approaches are scalable and adaptable to different farm sizes and landscapes. Farmers can deploy cameras strategically to cover large areas, monitor multiple locations simultaneously, and customize the system based on their specific needs and challenges.
4. Data-Driven Decision Making: The data collected through computer vision systems provide valuable insights into wildlife behavior patterns, enabling farmers to make informed decisions. By analyzing this data, farmers can identify trends, develop long-term strategies, and implement measures to optimize crop protection.
5. Conservation and Coexistence: By reducing crop damage, a computer vision-based approach fosters a more harmonious relationship between farmers and wildlife. It promotes coexistence and helps preserve biodiversity by minimizing the need for drastic measures that may harm or disrupt ecosystems.
6. This approach offers a non-lethal and sustainable solution to mitigate wildlife damage to crops.
• Computer vision enables farmers to implement targeted deterrence strategies, reducing the need for lethal measures.
• By preserving wildlife and ecosystems, the approach promotes long-term sustainability in agricultural practices.
7. The system improves resource allocation by accurately identifying the species responsible for crop damage.
• By pinpointing specific threats, farmers can focus their efforts on implementing measures tailored to address those species.
• This targeted approach leads to cost savings, reduced disturbance to non-problematic wildlife, and increased efficiency in resource utilization.
8. This approach is scalable and adaptable to various farm sizes and landscapes.
• Farmers can strategically deploy cameras to cover large areas and monitor multiple locations simultaneously.
• The flexibility of the technology allows customization to suit specific farm needs and challenges.
9. Data-driven decision making facilitated by it improves overall crop protection strategies.
• Analysing wildlife behaviour data helps identify patterns and trends, enabling informed decision making.
• This data-driven approach enhances the development of long-term strategies and optimizes crop protection measures.
10. A computer vision-based approach promotes coexistence between farmers and wildlife while preserving biodiversity.
• By reducing crop damage, the approach fosters a harmonious relationship between farmers and wildlife.
• It minimizes the need for drastic measures that may harm or disrupt ecosystems, ensuring the preservation of biodiversity.
Disadvantages of the Invention
While it offers many advantages for visually impaired people, there are also some disadvantages to consider:
1. High Initial Costs: Implementing a computer vision-based system can involve significant upfront costs. This includes the purchase and installation of cameras, and necessary infrastructure, as well as the investment in software development and maintenance. The initial investment may be a barrier for small-scale farmers with limited financial resources.
2. Limited Environmental Conditions: The system rely heavily on visual information, which can be affected by environmental conditions such as lighting, weather, and vegetation. Poor lighting or dense foliage may hinder the accuracy and reliability of the system, reducing its effectiveness in certain conditions or locations.
3. Limited Species Recognition: While computer vision algorithms can identify and classify certain species accurately, they may face challenges in recognizing less common or local wildlife species. The system's effectiveness may be limited to commonly encountered animals, potentially overlooking specific threats posed by less well-known or rare species.
4. Maintenance and Upkeep: The systems require regular maintenance and upkeep to ensure their optimal performance. This includes camera calibration, software updates, data management, and potential repairs.
5. Privacy Concerns: Deploying cameras for wildlife monitoring raises privacy concerns, as the system captures visual data that may also include neighboring properties or individuals. Adequate measures should be in place to address privacy concerns and comply with relevant regulations and laws governing data collection and storage.
, Claims:1. An AI and IOT based animal recognition, intrusion detection and repellent system comprises Battery (101), Camera(102), PIR Motion Sensor(103), Speakers (104), IOT Server (105) and Controller (Rasberry Pi) (106); wherein motion sensors and cameras is positioned in strategic locations to create a mannequin;
Characterized in that when an animal triggers the motion sensor, the camera will activate and use an AI computer vision algorithm to identify the species; Once identified, the system emits ultrasound emissions that are unique to each species to repel them; this intelligent sensing agriculture system is implemented using embedded edge AI and evaluated for its effectiveness in detecting and identifying various animal species; the system uses Convolutional Neural Networks (CNNs) to conduct the animal recognition process.
2. The system as claimed in claim 1, wherein the ultrasound doesn’t affect humans because animals have much greater sound sensitivity thresholds than do humans; they hear noises with lower frequencies than the human ear.
3. The system as claimed in claim 1, wherein if the frequency band is too high for the human ear, these ultrasounds do not cause any issues; human eardrums cannot vibrate under ultrasound frequency because they have a much lower specific resonance frequency than animal eardrums; PIR motion sensors consist of pyroelectric sensors that operate based on the principle of Pyroelectricity; pyroelectricity refers to the property of certain materials generating voltage when exposed to temperature changes, such as heating or cooling; every entity, whether living or non-living, emits a level of radiation, and the amount of radiation emitted is directly proportional to the entity's temperature.
4. The system as claimed in claim 1, wherein the PIR motion sensor consists of two slots made of a material sensitive to Infrared (IR) radiation, along with a lens; both slots can detect motion within their sensitivity range; in their idle state, when exposed to ambient conditions, both slots detect an equal amount of IR radiation; when an object such as a human or an animal approaches the sensor, it intercepts one half of the PIR sensor, causing a positive differential change between the two halves. Conversely, when the object leaves the sensing area, a negative differential change occurs.
5. The system as claimed in claim 1, wherein these positive and negative changes in differential trigger an output signal; this output signal is typically a pulse of approximately 3V in magnitude; in addition to the sensors, the lens on the PIR sensor plays a crucial role in its effective functioning; the lens surrounding the sensors is a Fresnel lens; the Fresnel lens design reduces the amount of material required compared to a conventional lens by dividing the lens into a set of concentric annular sections; once the PIR motion sensor detects motion, it triggers the activation of the Raspberry Pi's camera to capture an image; after capturing the image, the camera will record video for a duration of 5 seconds; the frames from this video is compared with a database using our suggested model
6. The system as claimed in claim 1, wherein if a matching animal is identified, the system initiates the release of ultrasound waves as a deterrent mechanism; the proposed Convolutional Neural Network (CNN) architecture consists of 1024 pixels, forming a 32x32 image; the CNN is divided into eight blocks, each serving a specific purpose in the network's operation; and provides a animal Recognition Model;
wherein the input data comprises animal faces from the dataset; to improve computation time, each animal face is resized to 32x32 pixels; the input database is expanded by scaling, rotating, and shifting the data to enhance experimental results;
wherein the second block represents a 2D CNN layer with 16 feature maps using a 3x3 kernel dimension, L2 regularization is employed due to the limited dataset, and the Rectifier Linear Unit (ReLU) serves as the activation function;
wherein in this layer, a 2x2 kernel is utilized, followed by dropout with a probability of 0.25, this step helps prevent overfitting of the neural network;
wherein the second 2D CNN layer employs the same parameters as the first one but with an increased number of feature maps, now totaling 32;
wherein a MaxPooling layer is applied, followed by dropout with the same probability;
wherein A standard dense layer with 256 neurons is used, and ReLU is applied as the activation function, L2 regularization is implemented for better weight control; and Dropout is set to 0.25;and the output dense layer consists of multiple classes, softmax activation function is utilized.
7. The system as claimed in claim 1, wherein in the proposed CNN architecture, pooling operations are applied separately to each feature map; the network repeats this process through successive layers, enabling the system to effectively recognize objects; for instance, the CNN may learn to detect edges in the first layer, use the edges to identify simple shapes in the second layer, and further utilize these shapes to recognize higher-level features like facial shapes in subsequent layers; in the CNN architecture, the neurons in each layer are organized in a 3D structure, allowing for the transformation of a 3D input into a 3D output; specifically, when dealing with image inputs, the first layer (input layer) represents the images as 3D inputs, where the dimensions correspond to the image's height, width, and color channels.
8. The system as claimed in claim 1, wherein the neurons in the first convolutional layer establish connections with specific regions of these input images and perform transformations, resulting in a 3D output; as the information propagates through the network, the hidden units (neurons) in each layer learn nonlinear combinations of the original inputs, thereby extracting relevant features; these learned features, also referred to as activations, serve as the inputs for the subsequent layer.
9. The system as claimed in claim 1, the process continues, with the learned features from one layer becoming the inputs for the next layer, until reaching the final layer of the network; the learned features obtained throughout the network ultimately function as inputs for the classifier or regression function, which make the final predictions or estimations.