A PORTABLE AI-DRIVEN CHILI GRADING MACHINE

Abstract

India is a major producer of chilies, with Andhra Pradesh and Telangana contributing significantly to national and global production. Post-harvest grading of chilies is essential but often hindered by labor shortages, high costs, and the need for timely processing. To address these challenges, we present a portable, AI-based chili grading machine designed for small-scale farmers. This device uses edge computing and artificial intelligence to automatically grade dried chilies based on color and quality. The process involves capturing images of chilies using a camera, labeling them by quality, and training a Convolutional Neural Network (CNN) model on cloud resources. The trained model is then optimized with TensorFlow Lite for efficient performance on edge devices like ARDUINO. Equipped with a camera and other necessary components, the device can continuously classify chilies in real-time, offering a cost-effective, labor-saving solution that allows farmers to quickly bring their graded chilies to market.

Specification

Description:FIELD OF THE INVENTION:
The present invention relates to the field of Artificial Intelligence in Agriculture. This invention is focusing on model of portable chili grading machine which sorts dried red chilies based on their color and quality. It integrates both edge computing and AI technology, making it portable and affordable. This device enables small-scale farmers to grade their chili crops with a low investment.
________________________________________
3. BACKGROUND OF THE INVENTION:
Chili crop is one of the most important commercial crops of India. The chili production in both Andhra Pradesh and Telangana is 57%, in India 39% and globally 69%. Post harvesting tasks in chili farming can be done in two phases called Drying and Grading. During the Grading process, farmers usually face problems like lack of labor, increase of labor cost based on demand, compulsion of labor work monitoring. Another important consideration is timely work completion. Then only farmers can reach to the market with in expected time and can sell their crop for good marketing price. Otherwise farmers may get loss due to floating of market price.

Manufacturers like Techik and Shenzhen Wesort Optoelectronic Co., Ltd. have developed a color sorting machines that are widely used in the industry. These machines are capable of handling large volumes of dried chilies. These machines are larger in size due to their functionality and design. For small-scale farmers, these advanced machines might be large and expensive.
Therefore, a portable and affordable AI based automatic chili grading machine is required for farmers to save their time, money, and dependency.
________________________________________
4. OBJECTIVES OF THE INVENTION:
1. Generate a proper dataset by collecting images of chilies using a camera which is same as attached to edge devices.
2. Build a CNN model and train the dataset using Cloud resources to classify chilies based on color and quality.
3. Optimize the trained model using TensorFlow Lite for deployment on edge devices.
4. Deploy the optimized model on edge devices like a ARDUINO with connected cameras.
________________________________________
5. SUMMARY OF THE INVENTION:
To create a portable and affordable grading machine, the proposed idea combines AI techniques with edge computing for colour classification of dried chilies. For classification, building a CNN model and training the model in the cloud leverages powerful computational resources, leading to faster training times and handle large datasets. This cloud-based training ensures that the model is robust and accurate, benefiting from the scalability and flexibility of cloud infrastructure. Optimizing the trained model using TensorFlow Lite for deployment on edge devices ensures that the model is lightweight and performs efficiently even on hardware with limited computational power. This optimization reduces latency and improves the responsiveness of the system, making it suitable for real-time applications. Because of all the above technologies, the machine will be available in affordable prize, portable size and with user-friendly environment.
________________________________________
6. DETAILED DESCRIPTION OF THE INVENTION:
Step 1: Collection of dried chili images and preparing dataset: The process of sorting dried red chilies starts from data collection. To generate proper dataset for chili classification/sorting, the chili images need to be collected using a camera which is same as cameras attached to edge devices to minimize the discrepancies in image quality. Label each image with appropriate class like A or B grade.
Step 2: Pre-processing of Dataset: The data pre-processing on collected data need to be performed to generate a consistent input data for model building. The dataset is then split into training, validation, and test sets.
Step 3: Model Building: Develop a model using suitable Convolutional Neural Networks (CNNs) architecture on the test dataset. During training, the model learns to recognize patterns and features in the images that correlate with the different classes of chilies. The Cloud resources can be utilized to train, validate the CNN model and experiment with different hyper parameters to optimize the model performance. Using cloud computing for processing tasks will minimize the need for large on board processors. It helps to make a portable machine.
Step 4: Optimizing model to make it compatible with edge computing devices: The trained CNN model is optimized using TensorFlow Lite converter to execute on edge devices. Because edge devices often have limited computational power and memory, so the model needs to be lightweight and efficient. So, TensorFlow Lite converter converts the trained model into a format that can be efficiently executed on edge devices.
Step 5: Deploy the model: The final step is to deploy the optimized model on edge devices. The edge device ARDUINO should be equipped with necessary peripherals like camera, power supply. The deployment process involves integrating the TensorFlow Lite model into the device's software environment. The cameras continuously capture images of chilies, which are then processed by the deployed model to classify them based on colour and quality.
, Claims:Claim 1: A model of designing a portable machine for chili grading system with AI and edge computing technologies consists of the following steps:
a) Collecting images of dried chilies using cameras ensuring consistent image quality across the dataset.
b) Pre-processing the dataset and build a CNN model on the train data.
c) Optimising the trained CNN model to execute on edge devices.
d) deploying the optimized model on edge devices to classify the dried chilies based on color and quality.
Claim 2: The method of claim 1, wherein the cloud resources can be utilized to train, validate the CNN model and experiment with different hyper parameters.
Claim 3: The method of claim 1, wherein using TensorFlow Lite, the CNN model is optimized to run efficiently on edge devices with limited computational power, ensuring smooth and real-time chili classification.
Claim 4: The method of claim 1, wherein the optimised CNN model deployed on edge device to capture images continuously, enabling real-time chili classification without relying on external computing resources.

Documents