Apostol, Kim Adams B. Domingo, Angelo C. Libunao, Christian Harold O. Sunga, Charles Edward T. Ticar, Gerald France A. Tiu, Joshua Migule Yaj A. 4 0
Philippine InsectiScan: an IoT and image processing device for identifying insects species for secondary science. 6 6 - - - - - - - - . - . - 0 . - . - 0 .
Undergraduate Thesis: (Bachelor of Science in Computer Engineering) - Pamantasan ng Lungsod ng Maynila, 2024.
5
ABSTRACT: STATEMENT OF THE PROBLEM: The Philippines, a country known for its rich biodiversity, faces numerous challenges in documenting, monitoring, and conserving its insect species. Many of these species are poorly understood, and traditional methods of taxonomic classification are time-consuming and require specialized expertise. Additionally, biodiversity assessments rely on outdated, labor-intensive approaches that limit the scope and accuracy of data collection. To address these issues, the project Philippine InsectiScan aims to develop an innovative solution that combines image processing and the Internet of Things (IoT) technologies. The key problem areas to be addressed are: . How precise will the system be able to distinguish the insects in terms of distance? . How accurate can the system be in classifying insects when operating in a day setting? . How reliable will the system be in classifying insects when operating in a night setting? RESEARCH METHODOLOGY: The study Philippine InsectiScan uses the like of IoT devices to collect real-time environmental data in ecological habitats and scanning images of insect species. The data is then processed using advanced computer vision algorithms for classification. The methodology involves monitoring emvironmental parameters and capturing images. The image processing phase involves preprocessing and machine learning algorithms for species identification. Field experiments will validate the methodology, involving collaboration with ecologists and taxonomists. The study also considers ethical considerations, minimizing environmental impact and responsible data handling. For the hardware component, a Single board computer (Raspberry Pi 4) and a High Quality Camera module (Rasberry Pi Camera Module V2 NoIR) were utilized in a custom-moduled 3D printed case which collaborates the two pieces together. The project was released gradually, with ongoing enhancements for improved usability. The software development followed a systematic approach, focusing on user requirements and needs, system accuracy, and intuitive interface design. The software and hardware teams worked closely together to ensure that the scanning of the insects and training of the datasets were in sync and make sure the integration went smoothly. Through user testing, live-testing and virtual image testing, Philippine InsectiScan was continuously improved based on training its dataset, becoming a dependable scanning service for students to use their studies. Data gathering involves acquiring pre-trained models and using a self-process acquiring of images required for the dataset. The system then deployed and went on a trial run on scanning both virtual and live subjects to further improve its feedback and studies. SUMMARY FINDINGS: The accuracy of the insect classification system varies based on the type of insect and the distance from the system, with notable differences observed daytime and nightimeevaluations. During daytime assessements, the system demonstrates accuracy, ranging from approximately 81.17% to 97.67% across various distances. Beetles (Coleoptera) exhibit the highest accuracy, followed closely by butterflies (Lepidoptera) and grasshoppers (Caelifera). However, moths (Lepidoptera) and spiders (Araneae) tend to have lower accuracy scores compared to other insect types. In nightime evaluations, the system's accuracy slightly decreases, with average accuracy ranging from 82.83% to 95.67%. Despite reduced visibility, beetles remain the most accurately classified insect type, followed by grasshoppers and butterflies. However, moths and spiders continue to present challenges in accurate classification under low-light conditions, Overall, the system demonstrates reliable performance in distinguishing insects based on proximity, with higher accuracy observed at closer distances. These findings underscore the system's adaptability and effectiveness in insect classification across varyingenvironmental conditions,. CONCLUSION: In summary, the insect classification system's accuracy is significantly influenced by the proximity of the insects to the system, with closer distances resulting in higher accuracy rates, gradually decreasing as distance increases. Nightime evaluations exhibit a modest reduction in accuracy compared to daytime assessments, indicating challenges in low-light conditions. Across diverse insect species, beetles (Coleoptera) demonstrate the highest accuracy, followed by butterflies (Lepidoptera) and grasshoppers (Caelifera), while moths (Lepidoptera) and spiders (araneae) exhibit lower accuracy levels. Despite variations, the system maintains a high level of overall reliability in insect identification and categorization, achieving reasonable accuracy rates across different illumination and distance scenarios. The insect detection and classification system performs well across a range of conditions, delivering dependable results at varying distances and illumination levels. However, there is room for improvement, particularly in enhancing accuracy during nightime operations and addressing challenges associated with speicifc insect species. Continued optimization and refinement efforts hold promise for enhancing the system's functionality and reliability in practical applications.
5
2 = =
2
2 --0------
6 --0-- 2 --------
0 2 --
--20------
--------20--
--------20--
----2
/ 2
/ 2
/
/
Philippine InsectiScan: an IoT and image processing device for identifying insects species for secondary science. 6 6 - - - - - - - - . - . - 0 . - . - 0 .
Undergraduate Thesis: (Bachelor of Science in Computer Engineering) - Pamantasan ng Lungsod ng Maynila, 2024.
5
ABSTRACT: STATEMENT OF THE PROBLEM: The Philippines, a country known for its rich biodiversity, faces numerous challenges in documenting, monitoring, and conserving its insect species. Many of these species are poorly understood, and traditional methods of taxonomic classification are time-consuming and require specialized expertise. Additionally, biodiversity assessments rely on outdated, labor-intensive approaches that limit the scope and accuracy of data collection. To address these issues, the project Philippine InsectiScan aims to develop an innovative solution that combines image processing and the Internet of Things (IoT) technologies. The key problem areas to be addressed are: . How precise will the system be able to distinguish the insects in terms of distance? . How accurate can the system be in classifying insects when operating in a day setting? . How reliable will the system be in classifying insects when operating in a night setting? RESEARCH METHODOLOGY: The study Philippine InsectiScan uses the like of IoT devices to collect real-time environmental data in ecological habitats and scanning images of insect species. The data is then processed using advanced computer vision algorithms for classification. The methodology involves monitoring emvironmental parameters and capturing images. The image processing phase involves preprocessing and machine learning algorithms for species identification. Field experiments will validate the methodology, involving collaboration with ecologists and taxonomists. The study also considers ethical considerations, minimizing environmental impact and responsible data handling. For the hardware component, a Single board computer (Raspberry Pi 4) and a High Quality Camera module (Rasberry Pi Camera Module V2 NoIR) were utilized in a custom-moduled 3D printed case which collaborates the two pieces together. The project was released gradually, with ongoing enhancements for improved usability. The software development followed a systematic approach, focusing on user requirements and needs, system accuracy, and intuitive interface design. The software and hardware teams worked closely together to ensure that the scanning of the insects and training of the datasets were in sync and make sure the integration went smoothly. Through user testing, live-testing and virtual image testing, Philippine InsectiScan was continuously improved based on training its dataset, becoming a dependable scanning service for students to use their studies. Data gathering involves acquiring pre-trained models and using a self-process acquiring of images required for the dataset. The system then deployed and went on a trial run on scanning both virtual and live subjects to further improve its feedback and studies. SUMMARY FINDINGS: The accuracy of the insect classification system varies based on the type of insect and the distance from the system, with notable differences observed daytime and nightimeevaluations. During daytime assessements, the system demonstrates accuracy, ranging from approximately 81.17% to 97.67% across various distances. Beetles (Coleoptera) exhibit the highest accuracy, followed closely by butterflies (Lepidoptera) and grasshoppers (Caelifera). However, moths (Lepidoptera) and spiders (Araneae) tend to have lower accuracy scores compared to other insect types. In nightime evaluations, the system's accuracy slightly decreases, with average accuracy ranging from 82.83% to 95.67%. Despite reduced visibility, beetles remain the most accurately classified insect type, followed by grasshoppers and butterflies. However, moths and spiders continue to present challenges in accurate classification under low-light conditions, Overall, the system demonstrates reliable performance in distinguishing insects based on proximity, with higher accuracy observed at closer distances. These findings underscore the system's adaptability and effectiveness in insect classification across varyingenvironmental conditions,. CONCLUSION: In summary, the insect classification system's accuracy is significantly influenced by the proximity of the insects to the system, with closer distances resulting in higher accuracy rates, gradually decreasing as distance increases. Nightime evaluations exhibit a modest reduction in accuracy compared to daytime assessments, indicating challenges in low-light conditions. Across diverse insect species, beetles (Coleoptera) demonstrate the highest accuracy, followed by butterflies (Lepidoptera) and grasshoppers (Caelifera), while moths (Lepidoptera) and spiders (araneae) exhibit lower accuracy levels. Despite variations, the system maintains a high level of overall reliability in insect identification and categorization, achieving reasonable accuracy rates across different illumination and distance scenarios. The insect detection and classification system performs well across a range of conditions, delivering dependable results at varying distances and illumination levels. However, there is room for improvement, particularly in enhancing accuracy during nightime operations and addressing challenges associated with speicifc insect species. Continued optimization and refinement efforts hold promise for enhancing the system's functionality and reliability in practical applications.
5
2 = =
2
2 --0------
6 --0-- 2 --------
0 2 --
--20------
--------20--
--------20--
----2
/ 2
/ 2
/
/