0 Introduction

Developing solutions to protect biodiversity in agricultural settings^[1] has become a key research topic as agriculture now occupies about half of the Earth's land surface^[2]. Many people in many parts of the world rely on agriculture as their primary source of income. Crop damage from animal and bird attacks is one of the leading causes of crop output reduction. Crop loss in agriculture due to birds is an ongoing and increasing cost to growers. Controlling birds is one of the main challenges farmers face every year. Crop attack by birds is one of the main reasons for crop failure and economic loss^[3-4]. Fig. 1 represents a schematic representation of autonomous eco monitoring system.

The amount of crop damage caused by birds is determined by a number of factors, including the bird population, cropping areas and patterns, the climate, and the birds' physical health^[1]. Bird damage in crops is more likely to reach an economic threshold where birds are numerous, feel at ease and have access to desirable crops^[5-6]. Damage is concentrated in area where bird habitat and susceptible crops adjoin. Orchard blocks and small fields with a higher edge area ratio are more susceptible, and the damage is worse when neighboring food sources are scarce and compromised. Controlling bird damage is more beneficial if you can keep them out long enough, they will find other food sources and may not return to your crop^[7-8].

Birds can wreak havoc on a wide range of crops, from fruits to grains. The need for treatments to reduce damage to agricultural fields caused by many cropping birds has been highlighted lately. Farmers lose harvest yield due to an increase in the number of cropping birds and their penchant for farming lands over wild foraging environments^[9] since crops give good quality food^[10-11].

1 Effect of Birds and Control Methods 1.1 Birds' Net Effects in Agro-Ecosystems

Birds are vital to science and society because of their powerful cascading effects on vertebrates, herbivorous insects, and plants, offering essential services including pest control^[12], seed dissemination, and pollination. Birds, on the other hand, can be a cause of agricultural damage. Many animals cause direct damage to the environment by granivory or frugivory, or indirectly through the consumption of natural pest foes, at a cost to growers. Concerns about birds carrying food-borne diseases are growing, prompting certain growers to decrease the access for birds to their fields. Other effects of avian habitation usage, such as predation of weed seeds and deposition of nutrients, could either benefit or harm agricultural systems^[13-14].

1.2 Pests Affecting Crop Field

Emerging plants and germinating seeds are damaged by pests like Invertebrates (e.g., flea beetles, slugs, maggots, wireworms) and vertebrates (e.g., birds, animals, etc.)^[15]. However, the type and extent of their harm are determined by the plot, field, or terrain parameters. Plot characteristics include size of plot, crop species, seeding date, and ploughing type; field characteristics include crop variation and natural or semi-natural habitations; and landscape characteristics include inhabited areas, forests, and annual vegetation^[16-17].

1.3 Bird Repeller

Crows, blackbirds, pigeons, starlings, sparrows, frequent myna and jungle myna, are the most common domestic birds in India as well as many other nations. Such birds will cause destruction to the cropping region as well as pollute the human living environment. Bird repellers are devices that use sound to drive birds away. A device called bird repeller is meant to drive birds away from newly planted arable crops^[18].

Electronic bird repellant systems create highly powerful acoustic and terrifying visual hazard, upset, also confuse birds, causing those birds to flee to safety^[19]. Table 1 shows the detailed procedures regarding the prevention methods for bird pest's control.

The bird repellent devices developed in the market mainly include wind-driven bird repeller, ultrasonic bird repeller, acoustic-optic bird repeller, and laser bird repeller, Sonic and ultrasonic bird repeller can work only for a short time because of the strong adaptability to sound wave of the birds. Due to its reliance on sunshine, wind-driven bird repellers have poor effects in the evening, cloudy, foggy, snowy and rainy days. Laser bird repeller acquires complicated energy with short service life and high rate of fault. High power laser bird repeller would interfere with the air route and bring hidden danger to the personnel climbing the line tower. Low power laser bird repeller yields poorer efficiency. Therefore, it is necessary to develop a bird-repellent fitting with simple structure, long service life, high reliability and efficiency for control of bird nests in transformer and main equipment framework^[20].

2 Studies on Bird Repellent System for Crop Protection

Khoomsab et al.^[21] described the use of chemical repellents to protect agricultural crops from bird predation. Anthraquinone is a stable chemical that is practically insoluble in water and has a minimal toxicity to birds and animals. The process of identifying and developing an effective and registered bird repellent chemical can be time-consuming, uncertain, and costly. The UV-Vis Spectro-photometer is a quick, easy and inexpensive way to detect the concentration of an analyte in solution. Many Thai herb species can repel birds and include compounds that can be used to make bird repellents.

Lindell et al.^[22] found that the significant thing in describing the damage to blueberries and grapes by birds is year, where, in high-yield years, the amounts of bird damage will be low. In high-yield years, bird management is not prioritized by the growers. Inflatable tube man and a methyl anthranilate spray are the two deterrence tactics, which will not reliably minimize the damage by birds. Two interesting bird deterrent tactics that should be researched further are drawing birds using nest boxes and disturbing birds' sensitively. Because birds are mobile and their activity varies across time and area, before-and-after sampling, as well as pairings of treatment and control blocks, should be addressed in deterrence experiments to enhance sample sizes.

Wang et al.^[23] proposed a new bird deterrence method involving numerous Unmanned Aerial Vehicles (UAVs) has been developed for vineyards. Agriculture-related bird harm is a serious and long-standing issue around the world. A successful bird-deterring system must be practical and autonomous to eliminate human operator costs.

Li et al^[24] proposed an intelligent bird repeller. First, two sophisticated deep learning networks are utilized for recognizing the species of birds, based on their appearance and vocalization. The bird repellent mode is then chosen based on the bird species. Finally, to scare the bird away, a special sound is played. The bird repellent methods are built using prevention performance analysis, and the sounds are bioacoustics, the adaption of birds to the bird repellent may delay.

Lindella^[25] investigated the effects of vertebrates on crop pests, fewer studies have examined how technological advancements could amplify trophic effects, which harm crops. In terms of crop pest effects, birds are the most studied vertebrates, whereas arthropods are the most studied pest group, and there have been several studies on coffee and chocolate. They lack knowledge on environmental and communal issues regarding developments, that includes 1) vertebrate predators which are most likely to be attracted towards developments and minimizing the agricultural pests, 2) the potential economic benefits of enhancements, and 3) how to organize human resources to fix, preserve, and monitor developments.

Azamjon et al.^[26] presented the notion of a bird repellent system, like a gadget that uses a stimulus to control wild birds. Many traditional measures are employed to deter these birds, such as colored lights, lasers, flash lights, scarecrows, hawk kites, chemicals, and so on, which are not efficient nowadays. The Solar Powered Audible Bird Scarcer was designed as an efficient bird deterrence approach in this study. Different noises that discourage different types of birds were also identified and investigated.

Andres et al.^[27] aimed to construct circuits to discharge electric shock and modify sound rate, create a program to track the appearance of the birds, manufacture the gadget, and test it in the field. The system used a Raspberry Pi 3, ultrasonic, and PIR sensors to automatically activate the set-up and log data. The device successfully deters the birds like pigeons, doves in the cultivated region after a series of tests. The T-test value of 10.25 is higher than t(0.05, 19)=2.093, which represents that the action was knowingly positive after applying to the test subjects, and a Pearson coefficient value of -0.95 indicates that more treatments were done to the diseased area. As a result, gadget can be used to keep birds away from farms as well as urban areas.

Bhusal et al.^[28] proposed a large multi-rotor UAS platform accompanied by a small quad-copter that was used as a technique for preventing activities of pest birds in wine grapes. No-flight and flight cycles were alternated for 14 days to evaluate the system performance in deterring birds. This technique showed that the number of birds coming into the field was lowered by 50% when UAVs were used in the field. Such level of reduction in bird count is expected to lead to a significant reduction in fruit damage caused by these pest birds. The use of large platform can be suitable to pose threat against birds in larger vineyards.

Klimek-Kopyra et al.^[29], with advancements in mechatronics, laser biotechnology can be used to stimulate plants in the field to reduce biotic and abiotic stress during their growth and development. This will increase crop yield and quality while lowering pesticide use, minimizing mineral fertilization, and enhancing crop plant nutrient use. Not only may laser biotechnology be employed in agriculture, but it can also be used in energy generation and environmental protection.

Wandrie et al.^[30] proposed a UAS to safeguard raw crop cultivation against bird pest damage. They looked into how captive and free-ranging red-winged blackbirds behaved to fixed-wing and rotary-wing (multi-rotor, quad copter) UASs by comparing preflight behaviors to behaviors during UAS approach. Due to the flight restrictions of the individual UASs, the fixed-wing and rotary-wing UASs were assessed at various heights. Blackbirds in captivity and in the wild showed no response, vigilance, or attempted escape/flight in response to approaching UAS.

Bapat et al.^[31] presented the development of an Agricultural Protection Wireless Sensor Network application for preventing animal infestations in the field of crops. PIR sensors, light flashers, RF modules and sound generators are installed in the agricultural field nodes. An incursion detection system is built at the farm's border to detect animals early. The nodes positioned at the farm boundary detect animal ingress and relay this information to the central base station.

Thirrunavukkarasu et al.^[32] proposed a mechanism to safeguard crops from animals and fire. This is a microcontroller-based system based on an Arduino Uno. A motion sensor is used to detect animals entering the field, and a smoke sensor is used to detect the fire. Table 2 presents overall comparative study on the Bird repellent system for crop.

3 Studies Based on Bird Repellent Using IoT Tchnology

Varghese et al.^[33] developed a cost-effective technology that, once implemented, will provide real-time agricultural information. The technology uses IoT and machine learning to create a cost-effective smart farming module. This technology employs cutting-edge techniques to improve the accuracy of results and automate crop monitoring, needing little human participation. The IoT is utilized to link the ground module with the sensors to the cloud infrastructure. Machine learning-based real-time analytics are used in the cloud to forecast future crop conditions based on historical data.

Chen et al.^[34] proposed a scheme for constructing an intelligent bird repelling system in an airport using IoT technology, as well as a method for an automatic bird blocking network, it addressed the issue that physical bird-blocking networks cannot be combined with the airport intelligent system. At last, the research showed that the constructed automatic bird-blocking net performs netting and netting functions like typical bird-blocking nets while also increasing the automation degree of the real field bird-driving tools.

Boursianis et al.^[35] described the essential standards of IoT technology, such as intelligent sensors, IoT sensor types, networks and procedures utilized in agricultural science, and IoT implementations and results in smart agriculture^{[9, 36]}. Furthermore, the relevance of UAV technique in intelligent farming was demonstrated by examining UAV implementation in a variety of situations, including irrigation, fertilization, pesticide use, weed control, plant growth observation, agricultural crop disease control, and field-level phenotyping. In addition, the use of UAV techniques in difficult cultivating situations is investigated. The result is that the IoT and UAV are the most essential techniques, which are transforming customary agricultural techniques to a different level of intellect in accurate agriculture.

Ramadhan et al.^[37] discussed that farmers confront a variety of challenges as a result of the presence of bird pests, which can result in a reduction in the standard and number of birds. An effective method for bird watching and management in the field of rice is the use of WSN. PIR sensors are the most common device for sensing bird pests, whereas buzzers are used to repel bird pests. Mesh topology is utilized in the construction of system, then each sensor may communicate in two directions and know about situations of others.

Roihan et al.^[38] examined many ways of bird detection using video sensors and selected the most accurate method, which was then used to automatically repel birds using sound frequencies. It was despised by birds and deterred the bird pests. The Bird Repellent system was designed to collect bird objects in each frame using computer vision techniques via camera sensors, which were subsequently processed by a microprocessor. The actuator will be stimulated by the microcontroller via sound frequency after the object has been caught on the camera. The purpose of the project, which is based on the Internet of Things, is to create procedures for the development of observing and autonomous mechanisms of bird attacks in order to increase agricultural crop yields

Chen et al.^[39] discussed the gas gun concept for the intelligent bird-driving system. The birds swiftly "developed" the inertial signal, which was nearly "immune" to the tools, as a result of the distinct, uninteresting, and purposeless method to bird driving in the present airport field, putting tremendous pressure on the airport's work. With the growing number of flights, current research on driving birds has shifted to how to create and improve an intelligent system for driving birds, in order to make driving birds more scientific, intelligent, and effective. The project's intelligent bird drive system integrates existing Internet of things technology with the motion of the bird drive tools to effectively minimize blind startup of the bird driving tools due to human errors.

Riya et al.^[40] proposed the model and prototype of an automated bird detection and repeller system using IoT devices. The annual income of farmers largely depends on the yield of crops that they produce, which is continuously decreasing due to a number of factors. One such factor that we are focusing on is the damage caused by birds. The bird repellent system proposed by Riya^[40] consists of two main functionalities, motion detection using PIR (Passive Infrared) based motion detectors, and a repeller that generates sounds of predators to drift the birds away from the field, using an MP3 module and megaphone.

Abdellah et al.^[41] presented a user-friendly Internet of Things (IoT) solution to assist farmers, particularly in rural areas, in remotely visualizing their agricultural data, resulting in time savings, increased crop output, and precise irrigation. This technique makes it possible to lessen the number of farmers abandoning agriculture for mining. The design is created by connecting the actual equipment in the agricultural field to the operator mobile application using the Blynk IoT platform, allowing the farmer to visualize the data. The Raspberry Pi 3 serves as the controller for all operations, including data transmission and reception using humidity and temperature sensors, soil humidity sensors, pH sensors, PIR, and video sensors, as well as a water pump as an actuator.

Manickam et al.^[42] proposed a smart cultivation scheme that includes soil sampling, seeding through drone, insecticide/fertilizer spewing, and condition of agricultural field is monitored by the use of IoT sensors and drones. The agricultural data such as soil moisture and pH value from a ground sensor is collected using the drone that will be pre-plugged in the agricultural land by the farmer. Because of its capacity to connect over a long distance, Zigbee is used to connect the ground sensor and the drone. The acquired data will subsequently be transmitted through LoRa to a gateway, where it will be stored for data storage and analysis in the cloud.

Kalra et al.^[43] developed a water system architecture that enhances the availability of water in the water source, thereby providing an active and dominant method for water supply. When the moisture sensor detects the amount of water in the reservoir, the water supply system would habitually start/stop water drain off on the cultivating region based on the humidity content. The Arduino Uno microcontroller receives the calculated sensor estimates and arranges the calculation. The crops are protected from the pests using voice detection and movement detection systems.

Chen et al.^[44] proposed a method of constructing an airport bird-repelling linkage system using the IoT technology. Because the Zigbee network has flaws and is constrained by the peculiarities of the airport, a solution to the issue of communication distance between the bird repellant tool and bird control center is required. In an intelligent bird-repelling system, the ZigBee-GPRS gateway will provide remote control, real-time communication, and observation between bird-repelling tools and workers.

Balaji et al.^[45] proposed a system for agricultural field crop monitoring that is both efficient and effective. Data may be stored and retrieved from anywhere via the Internet of Things. The sensor element of this suggested work is confined to crop monitoring; but, in the future, it might be automated for irrigation, and the system may be augmented with field security under video surveillance to prevent infiltration.

Chourey et al.^[46] designed a smart agricultural protection system to solve the problems of farmers with the use of IoT. The main goal is to reduce corpse loss and to safeguard the area from wild animals that inflict significant damage to the agricultural sector. As a result, the technical approach will assist agriculturalists in safeguarding the field crops and save them from economic losses as well as unproductive efforts that they must bear in order to protect their farms.

Srivastava et al.^[47] described the method for Man-Animal conflict by using a laser fencing system which was based on IoT Technology. The system detected the presence of animals and alerted the owner of the field. The system was working through ESP-8266. The sound creating devices and high intensity lights were used to distract the animals from the field and the day-night vision camera was used to capture the image.

Giordano et al.^[48] presented the development of a crop protection IoT implementation for avoiding animal invasions on the farm. To mitigate probable harm by both wild animal attacks and meteorological circumstances in agricultural field, a deterring and observing system was given.

Prathibha et al.^[49] employed CC3200 single chip sensors to monitor temperature and humidity in an agricultural area. In smart agriculture, the Internet of Things (IoT) plays a critical role. IoT sensors are proficient for supplying data on their farming fields, so smart farming is a new concept. The camera is connected to the CC3200, which captures images and sends them through MMS to the farmers' cell phone via Wi-Fi.

Upadhyay et al.^[50] designed a system that detects livestock movement in the fields using infrared passive infrared sensors. These sensors are placed at each of the agricultural field's four corners. The sensor sends the data to the transmitting part, which then sends it to the cloud. On his cell phone, the farmer gets a message with the alert through an internet connection. The message informs the recipient of the cattle breaching zone. Table 3 provides the comparative study on Bird repellent using IOT technology.

4 Studies Related to Crop Protection from Pest Birds Through UAV

Ahmad et al.^[51] presented the study of methylanthranilate and anthraquinone, two bird repellents. In aviary settings, they are compared to house sparrows on maize seeds and seedlings. The experimental group in aviary-I received varied doses of both bird repellents on maize seeds and saplings, whereas the control group in aviary-II received untreated seeds and plants for three hours, early in the morning. Two closed circuit cameras were mounted in each aviary to track the behavioral responses to various doses of both chemical repellents. The experimental and control groups for seeds and plants exhibited highly significant differences in statistical analysis. Significant differences in efficacy were found when comparing the two repellents, with anthraquinone outperforming methylanthranilate in maize seedlings.

Mohamed et al.^[52] proposed an inexpensive, harmless bird frightening approach to frighten the birds from the rice fields. Finally, combining a visual and acoustic system on a drone results in a different vibrant method for bird management. The birds' temporary reactions suggest that the drone is used to frighten the birds, however it would have to be done repeatedly. Ground vehicle interferences were discovered to have an instantaneous effect on scaring birds in paddy fields.

Olimpi et al.^[53] found that birds' net effects on strawberry yield change in complex but predictable ways, owing to trophic relationships and farming setting. Despite the fact that birds' net effects were somewhat unfavorable overall, the study shows that semi-natural environment is able to support to attenuate bird damage. These results can provide general management advice that improves bird facilities while decreasing bird damage, and they may be relevant to various agricultural systems and geographical situations when considered in conjunction with how producers manage hazards associated with fecal contamination.

Wang et al.^[54] suggested a revolutionary Unmanned Aerial Vehicle, called as a drone method for effective bird damage management that incorporates bird psychology. The UAV has a speaker that broadcasts anguish signals, along with taxidermy crows on the underside that seem like trapped prey. This particular UAV design is intended for participating birds' fixed abilities to recognize and reduce new animal of prey, with the added use of a durable anxiety reaction to the UAV. The results showed that the UAV will discourage huge problem of birds like bolts and cockatoos for a lengthy period of time in a 50 m radius centered on the UAV. Some of the pest birds like silvereyes, can be efficiently deterred by the UAV for short periods of time after being exposed to it. The results also showed that, numerous UAVs were required for properly safeguarding a big vineyard, while one UAV was adequate to guard vineyards less than 25 ha.

Amanda et al.^[55], to prevent blackbirds from objective areas of mature sunflower fields, used a unique deterrent: sound meant to disguise message between birds (dubbed a "Sonic Net"). The Sonic Net hides an objective species' communication by sending "pink noise" which overlays by the frequencies used for auditory communication. When birds can't hear predators or conspecific warning sounds, they perceive a higher risk of predation and flee to a safer place.

Chan et al.^[56] presented a crop security system capable of automatically identifying and notifying intruders. Green electrical energy created by both a solar board and a wind generator is employed to ensure that the system has enough electrical power. Five ultrasonic sensors, a microprocessor, twinkle lighting, and a speaker unit are included in this system, which is installed in each region of the farm. Crops in large areas will be separated into multiple small zones for monitoring in order to ensure efficiency when employing the crop security system. Every zone will have its own crop protection system. When a sensor senses a bird invasion, an alarm and light will go off. Any birds or other intruders will be scared and leave the field as a result of this.

Although the ultrasonic waves have a limited area of impact, Arun et al.^[57] gradually frighten the pest birds from the selected areas by creating a hostile environment for them and having a repelling effect on them, The weaver and black birds, but not quelea birds, exhibited a visible reply to an ultrasonic wave stimulation emitted by ecologically responsive device. The waves travelled further as the device's power increased, and on wet days, the waves travelled further than on dry days. This is helpful since rain-fed cereal crops bear fruit during the rainy season, necessitating the use of the device at that time.

Marcoň et al.^[58] constructed and categorized an AI-based method for detecting ripe fruit-damaging bird flocks in real time. They fixed the assurance ratio to 30% in order to address no error recognition. The microcomputer's algorithm will activate the actuator wirelessly at increased rates, triggering the bird scaring procedure. CNN and Artificial Intelligence are used to conduct the actual detection. For detecting every moving object in the vineyard before training the network, video cameras and a differential algorithm were used. The images' exposed objects were identified and used in the network's training, testing, and validation.

Adebayo Segun et al.^[59] presented the vision-controlled quadcopter device, which is used to find and chase birds in the planted areas. The strategy entails utilizing robot perception^[60] to manage the quadcopter's location by an effort to pursue the item during frightening sound emissions such as bird anguish calls and predator calls.

Goel et al.^[61] constructed a fully autonomous Unmanned Aerial System (UAS), a bird detection and finding system, to protect birds from blueberry and grape vines. The vision system is the most important aspect of the UAS's implementation in the attempt to construct it. The algorithms for background removal were employed to detect birds, and the performance of different background subtraction techniques has been tested. ViBe, a background removal algorithm, was found to perform best in the bird detection situation, with a 63% accuracy. A split window technique was utilized, which raised the detector speed by 13% in order to improve the bird's speed and gather it in real time.

Wang et al.^[62] presented a method for utilizing teams of unmanned aerial vehicles (UAVs) to discourage autonomous birds. Birds wreak havoc on commercial crops all around the world. The limitations of conventional bird control approaches could be overcome by a bird deterrent system using autonomous UAVs. A model predictive control issue was used to formulate the problem of controlling UAVs to repel birds. The occupancy gird map is used to create a world model that represents the system's knowledge of target birds' current locations.

Nair et al.^[63] developed a technique to safeguard the field against livestock and birds that enter the field to harm yields that are ready to harvest. For this, a fence around the agricultural land should be implemented, and this fence should be powered by solar energy, so that whenever an animal touches the field, they should receive a short pulse that does not injure them, but threatens to keep them out of the field, as well as an ultrasonic bird repellent that transmits ultrasonic waves so that birds do not get into the field and destroy the cultivations. To water plants/crops, an automatic water supply system will be employed, which will operate by sensing the soil moisture level and watering plants / crops. Farmers will benefit from the development of such a system since it will shift manual monitoring to system level monitoring and educate farmers about system operations in farmland via GSM.

Dayoub et al.^[64] argued that every species or group of animals has a unique hearing frequency range. A special logic was used to estimate the irritating frequency. Therefore, creating irritating sounds can make the birds fly outside of the field. By using this research idea mostly affected problem in agriculture can be reduced.

Borah et al.^[65] introduced a microcontroller-based sound protection system to safeguard farmland from crop vandalism by creating a sound sensor for a band-pass filter (BPF). Currently, the majority of our farmers use traditional methods to protect their crops against bird attacks. However, these methods are ineffectual and time-consuming. This study seeks to give an automatic monitoring system to our agriculturists in order to handle the above-mentioned problem in a cost-effective and efficient manner. Comparative studies on crop protection from pest birds using UAVs are presented in Table 4.

5 Studies Related to Birds Sound Detection Using Deep Learning

Solomes et al.^[66] proposed the Bela embedded Linux device's automatic detecting method, which underlines the device's user friendliness of development and an extremely low latency audio processing environment that places audio before other operating systems. On a commercially available platform, the algorithm delivers high-quality recognition while running efficiently on continually streamed data. The computer was able to calculate the on-board detection utilizing convolutional neural networks (CNNs) with an 82.5% AUC score when testing an international data challenge.

Santosh Bhusal et al.^[67] presented the use of super-resolution technology to improve the small moving object's quality that was afterwards categorized as birds or false positives with the use of deep learning. The use of super-resolution improved image resolution, results in a higher density of pixels and better scene detail. The CNN-based classifier acquired better feature information as a result of super-resolution, allowing it to make more educated decisions when classifying birds. After resolution augmentation, the classification accuracy increased significantly from 70% to more than 90%.

Oisin Mac Aodha et al.^[68]developed open-source pipeline using convolutional neural networks to locate ultrasonic, full-spectrum search noises by echo-localizing bats. Deep Learning algorithms were taught with ultrasound full spectrum captured throughout Europe on road transport and labelled by citizen scientists. In the test sets the search-phase echolocation sounds demonstrated much higher performance detection than other existing algorithms and commercial systems. As an example, researchers have used data obtained in Jersey (UK) over five years to test using their detection process and compared the findings to a well-used profitable system. This recognition process is used to detect and monitor bat populations on a wide scale, making them more useful as indicator species. The suggested pipeline only makes a few bat-specific design decisions, and it may be used to recognize other species in audio given enough training data.

Yahot Siahaan et al.^[69] established a bird repellant and prototype detector. The model uses a moving object detecting method, which triggers a repellant that emits a specific frequency, causing the bird to be agitated. PIR sensors are utilized as bird motion sensors and LC oscillator Colpitts are employed as repellents using an Ultrasonic Sensor. When birds are recognized, ultrasonic pulses are automatically provided to the birds.

Oluwole Arowolo et al.^[70] proposed an effective repellent system, which is made up of hardware components that include the raspberry pi for image processing, the servo motors for rotation of camera for better field of view controlled by Arduino connected to the raspberry pi, a speaker for generating predator sounds to scare birds away and software component consisting of python and Open CV library for bird feature identification.

Dan Stowell et al.^[71] presented new data sets for acoustic monitoring, a review of the machine learning algorithms proposed by challenge teams, a comprehensive assessment of performance and a discussion on how these methods might be used in remote monitoring projects. Multiple technologies achieved an area of around 88% under the Receiver Operating Characteristic (ROC) curve, substantially higher than previous general-purpose procedures. General acoustic bird detection can achieve very high recovery rates in remote data monitoring employing state-of-the-art machine learning, including profound learning without manual reassessment and no detector pre-training for target species or auditory parameters in target environment.

Seolhee Lee et al.^[72] suggested an adaptive deep learning to certain cultivated areas where the possibility that the bird will occur is high, based on image processing results. The motional objects are first retrieved using a Gaussian Mixture Model to conduct background subtraction. The superfluous elements in the agricultural landscape are then removed using color extraction and a median filter. In the neural network object classification, the reduced moving objects are classified.

Thomas Grill et al.^[73] developed two approaches to detect the existence of audio bird sounds using convolutional neural networks on mel spectrograms. In a signal processing challenge using ambient recordings from three very different sources, they received an Area under Curve (AUC) value of 89% on the concealed test set, which was higher than any other competitor, and only two of which were offered for supervised training. They discovered that, although having quite different structures, both methods can be twisted to function similarly fine by comparing various iterations of the systems.

Yang et al.^[74] designed a wireless system for airport bird monitoring, finished the lower computer's analogue attainment and data handling, completed the lower computer's and upper computer's ZigBee module setup, and completed the upper computer's command transmission and data display. Signals are collected through an STM32 embedded control and data processing core by the temperature sensor, pressure sensor, power supply tension and sound sensor of bird repelling equipment.

Cheol Won Lee et al.^[75] proposed an Anti-adaptive Harmful Birds Repelling (AHBR) method, which was used for repelling harmful birds that employs the Reinforcement Learning (RL) approach's model-free learning concept to efficiently avoid bird accommodation difficulties. To avoid habitation, the AHBR method employs a technique that involves studying the bird's behavior to accessible dangerous noises and play in patterns that are hard to adjust using the RL approach. The Long-term and Short-term (LaS) strategy was also proposed to address the Markov conventions which make RL hard to execute. The LaS strategy allows researchers to learn about a bird's true reaction to a danger sound.

Li et al.^[76] created a whitefly and thrips detection algorithm using adhesive trap images collected in greenhouse conditions. To increase the accuracy of micro pest detection, faster regional-convolutional neural network (R-CNN) is constructed based on an end-to-end model, dubbed TPest-RCNN. The TPest-RCNN model was trained on the collective things in environment dataset using a transfer learning method before being trained on the micro pest training set.

Song et al.^[77] designed a Binarized Convolutional Neural Network, which is used to create an audio categorization algorithm for BSD (BCNN). The convolutional and fully connected layers of the original Convolutional Neural Network were binarized into two values. In an unseen examination, the BCNN's Area Under ROC Curve (AUC) score attained equivalent performance to the CNN. This study showcases two different networks (CNNs and BCNNs) for the BSD problem of the IEEE AASP Challenge on Acoustic Scenes Detection and Classification of Events (DCASE2018). BCNN's ROC (AUC) area was equivalent to CNN's on unseen evaluation data.

A CNN approach for categorizing bird noises was described and evaluated using various configurations and hyper-parameters by Agnes^[78]. A dataset obtained from the Xeno-canto bird song sharing portal, which contains a vast group of labelled and considered recordings, is used to fine-tune the MobileNet pre-trained CNN model. The neural network's input is represented via spectrograms produced from the downloaded data.

Devika Sunil et al.^[79] noticed that Agriculturalists are not permitted to surround whole fields with barricades or to keep them harmless for 24 h a day. A spontaneous method for bird protection was proposed. This method includes cameras monitoring the field 24 h a day, seven days a week, and ultrasonic sound waves are used to annoy the birds when they are discovered within the field. Deep Learning technology has advanced to the point that it can be used in this industry, lowering the amount of manual effort required for region surveillance. Object recognition in video streams is done using the YOLO method.

Mario Lasseck^[80] proposed deep learning algorithms for detecting acoustic birds. Originally intended for image classification, Deep Convolutional Neural Networks (DCNNs) have been repurposed and fine adjusted to detect birds' presence in audio captures. To improve ideal presentation and generalization to indefinite recording settings and environments, various data augmentation techniques are used. The suggested method is tested on the dataset from the IEEE AASP test on Detection and Classification of Acoustic Scenes and Events (DCASE). Table 5 depicts detailed analysis of different Deep Learning approaches for birds sound detection.

6 Proposed Methodology

Here is the proposed methodology for a bird repellent system with deep neural networks:

1.Data collection. The first step is to collect data of bird species and their calls. This can be done by recording bird calls in the field or using a library of pre-recorded bird calls. The data should be labeled with the species of bird that is calling.

2.Data preparation. The data collected in the previous step needs to be prepared for training the deep neural network. This includes normalizing the data, removing noise, and splitting the data into training, validation, and test sets.

3.Model selection. There are many different deep neural networks that can be used for bird classification. The most appropriate model will depend on the specific application. Some popular models for bird classification include convolutional neural networks (CNNs) and recurrent neural networks (RNNs).

4.Model training. The deep neural network is trained on the labeled data collected in the previous step. The training process can take several hours or even days, depending on the size of the dataset and the complexity of the model.

5.Model evaluation. Once the deep neural network is trained, it is evaluated on the validation dataset. The evaluation results can be used to fine-tune the model or to select a different model.

6.Model deployment. The trained deep neural network can be deployed to a device that can detect bird calls. This could be a standalone device or a smartphone app.

Here are some additional considerations for implementing a bird repellent system with deep neural networks:

·The accuracy of the system will depend on the quality of the data used to train the deep neural network.

·The system should be able to detect a wide variety of bird species.

·The system should be able to distinguish between bird calls and other sounds, such as human speech or traffic noise.

·The system should be able to emit a sound that is aversive to birds but not to humans.

7 Conclusions

Farmers are greatly harmed by the large number of birds attacking in colonies or groups. In this paper, different bird repellent methods based on IoT, Unmanned Aerial System and Deep learning technologies are reviewed. No bird repellents will provide 100% safe guard. The IoT technology is the most significant contribution of research since it will lead to a new area of research. Also, we have noted some methods which give accuracy but have high computational complexity. Specifically, the methods implemented to increase the performance of the system are discussed.