Entry No. 002
Junior (under 14)/ Senior (14 and above)
Senior (14 and above)
Category
6.2.1.1(6) Visionary Technologies
Classification
C. combination
Country / Region Thailand
Name of Invention/Artwork The Prototype of AI-Powered Rock Salt Crystal Classifier
Outline of the entry
The aims of this study were: (1) to develop a prototype machine capable of classifying rock salt crystals using artificial intelligence (AI), and (2) to evaluate and compare the performance of the AI-powered rock salt crystal classifier with traditional manual sorting methods. The use of AI aimed to enhance the efficiency of rock salt crystal classification and provide a new model for increasing the income of local communities.The results revealed that the prototype AI-powered classifier could distinguish between two types of rock salt crystals: perfect crystals and imperfect crystals. The prototype demonstrated an 80% accuracy rate in classifying perfect rock salt crystals. However, this performance was limited by incomplete AI training and issues related to lighting and shadow, which occasionally caused detection errors by the camera system. Additionally, the prototype achieved a 60% accuracy rate in identifying imperfect rock salt crystals. The lower accuracy was attributed to the high visual similarity between perfect rock salt crystals and imperfect rock salt crystals, which led to misclassification by the detection system. The results obtained are solely based on the prototype of the AI-Powered Rock Salt Crystal Classifier and cannot yet be applied for industrial-scale use. The development team will use the identified issues as a basis for improvements in the next version of the AI-Powered Rock Salt Crystal Classifier.
Characteristics of the entry
Bokluea salt ponds are a unique tourist attraction, known for the traditional process of boiling salt, which serves as a source of income for the local community. The salt production process produces two types of salt: rock salt (Sinthao Salt) and rock salt crystals (Sinthao Salt Crystals).When compared to rock salt, the rock salt crystals are priced higher due to their more complex production process. Additionally, when used in cooking, rock salt crystals enhance the flavor of food, making it more flavorful and aromatic. The price of rock salt typically ranges from 20-25 baht per kilogram, while rock salt crystals are sold for 100-300 baht per kilogram, depending on the purity and quality of the crystals. The higher the purity, the higher the price.
In the community, the traditional method for sorting the salt crystals involves using a wide-mesh sieve to sift out the smaller pieces, followed by manual inspection to separate the remaining crystals. This process is time-consuming and, if not done skillfully, can result in the crystals breaking or being damaged.Our group came up with the idea to develop a prototype AI-powered salt crystal sorter to improve the efficiency of the sorting process. This machine helps save time and reduces the damage that occurs during manual sorting compared to traditional methods. By producing more perfect rock salt crystals, this innovation will increase the income of local producers and vendors in the community.
Demonstration details
The device serves to classifying perfect rock salt crystal and imperfect rock salt crystal by Initially, the rock salt crystals are placed onto the conveyor belt. After that, the system is activated by pressing the switch, causing the machine to operate.As the conveyor belt moves, the rock salt crystals are carried past a Web Cam equipped with artificial intelligence (AI), which analyzes and assesses the quality of each crystal.The quality data detected by the AI is then transmitted to a Micro:bit microcontroller, which processes the information in real-time.Once the crystals reach the sorting section, the Micro:bit sends a command to the Servo Motor to rotate in a specified direction. The rotation mechanism separates the crystals into different categories based on their assessed quality.
Other notes about the entry (if any)
Identified challenges:
1. The AI training process was not fully optimized, which impacted overall classification accuracy.
2. Variations in lighting and shadow conditions interfered with the camera’s detection capabilities, leading to occasional misclassifications.
3. The physical similarities between perfect and imperfect crystals further contributed to the difficulty in achieving precise classification.
The results obtained are solely based on the prototype of the AI-Powered Rock Salt Crystal Classifier and cannot yet be applied for industrial-scale use. The development team will use the identified issues as a basis for improvements in the next version of the AI-Powered Rock Salt Crystal Classifier.
Information on patent, utility model, trademark, etc. application
Number of Team Members
2
Student
Miss Thamonwan Khanluang March 9, 2007 tbaikhao@gmail.com
Master Kunakorn Bualek July 22, 2008 kunakorn.bualek@gmail.com