Autonomous manipulation platform for tissue culture operation

Project

• Designed and integrated a hardware setup with RaspberryPi, RealSense camera, and manipulator arm; implemented ROS 2 MoveIt2 for autonomous transfer of callus during tissue culture operation,
• Developed callus detection algorithms, camera calibration, and collision safety controls using point cloud data to enhance manipulator accuracy.

A Two-stage Deep-learning Model for Detection and Occlusion-based Classification of Kashmiri Orchard Apples for Robotic Harvesting

Divya Rathore, Divyanth L G, Lalit Reddy, Yogesh Chawla, Mridula Buragohain

Paper

• Proposed a novel two-stage deep-learning-based approach that can detect the apples using YOLOv7 and utilized EfficientNet to classify the apple’s occlusion condition.

Classification of tea according to the texture and color analysis based on origin using computer vision

Dhanus Raj, Ingrid Moraes, Mridula Buragohain, Sai Chaitanya K, Vansh Agrawal, Vanshika Sahu, Ayushi Saha, Marcus da Silva Ferreira, Sylvio Barbon Junior, Somsubhra Chakraborty, Jayeeta Mitra, Douglas Fernandes Barbin

Conference Presentation

• Developed machine learning models, including SVM, Random Forest, and ANNs, for classification based on using color, texture, and morphological features extracted from RGB images. Achieved a peak prediction accuracy of 89% with the Random Forest classifier.

Adaptability and Effectiveness of inbuilt vision foundational model to quantify grape foliage powdery mildew infection

Mridula Buragohain, Yu Jiang, Yiyuan Lin

Poster Presentation

• Tested our built vision foundation model for adaptability results across different vineyards in the US
• Obtained bird-eye view severity assessment results of mildew infection in the fields compared with human assessment.

Mechanical damage assessment in canola using X-ray imaging and computer vision

Summer Project

• Utilized X-ray imaging with computer vision techniques, including OpenCV for image enhancement and ROI detection, to classify canola damage into four categories. Tested machine learning and deep learning models, achieving a top mAP of 92.08% with MobileNetV2.

Simulating an autonomous navigation system for a ground robot using NAV2

Semester Project

• Developed and tested a ground robot in the Gazebo simulation environment for autonomous navigation and obstacle avoidance using NAV2, Lidar, and the A* algorithm. Gained hands-on experience with ROS2 for controlling both virtual and real-world robot designs.

Quality analysis of mangoes during storage using color imaging

Project

• Conducted quality assessments of mangoes during storage by applying classification algorithms on RGB and lab images, comparing machine-vision results with chemical-based methods. Achieved an 84% accuracy in predicting the ripening index of the mangoes.