VISITA DEL SINDACO DI TRENTO PRESSO IL LABORATORIO MIRO 24 Maggio 2021
Con grande piacere nella giornata odierna Lunedì 24 Maggio abbiamo accolto il Sindaco Dott. Franco Ianeselli presso il nostro laboratorio di Misure Meccaniche e Termiche del Dipartimento di Ingegneria Industriale dell’Università di Trento. Dopo una breve introduzione del laboratorio da parte del Prof. Mariolino De Cecco, una delegazione di dottorandi e ricercatori tra cui Alessandro Luchetti, Nicola Covre e Matteo Zanetti ha presentato alcuni dei progetti ai quali stanno lavorando. Nello specifico il progetto Ausilia in collaborazione con APSS, i progetti Europei Eurobench, Mirebooks, la collaborazione industriale con il gruppo GPI ed internazionale con il centro di ricerca giapponese NAIST. In Figura al centro il Sindaco Dott. Franco Ianeselli ed il Direttore del Dipartimento di Ingegneria Industriale il Prof. Dario Petri, al lato il responsabile del laboratorio Miro il Prof. Mariolino De Cecco, il prof. Paolo Bosetti ed alcuni tra dottorandi, ricercatori e Professori del Dipartimento. Continue reading →
AUSILIA – General Idea 24 Maggio 2021
The project was developed by UNITN and APSS at the clinical rehabilitation hospital of Villa Rosa (provincial reference for rehab medicine).
The project AUSILIA (Assisted Unit for Simulating Independent LIving Activities) aim at the personalization of the technological and architectural design in order to Abilitate impaired users independency and reduce re-hospitalization burden.Continue reading →
Designed and Developed by Nicola Covre
Special thanks for the support of Paolo Tomasin and Andrea Vigorito
Designed and Developed during the Covid-19 Quarantine, March-May 2020
The proposed system was built using simple Arduino-compatible components and recycled packaging material.
The system consists of 2 Arduino Leonardo, mounted respectively on board the aircraft and in the radio control, 2 NRF24 radio communication units, 4 servomotors, 1 11.1V DC motor, 2 propellers glued together, 1 11.1V Lipo battery, 1 Esc30A.
The structure is mainly made of cardboard with a wooden support frame and a tip protection made of a tennis ball.
The transmission of control to the moving surfaces was made with rigid rods by recycling paper clips.
Continue reading →
Research on Feature Descriptors used for Point Cloud Registration
Team SLAMer: Ma Teng, Sharad Maheshwari, Mingxuan Líu
Supervised by Alessandro Luchetti & Prof. Mariolino De Cecco
1. Problem description:
In order to achieve SLAM, first we need to get the information of the environment, here we use a 3d camera to generate PCD files in unity (find more in appendix). Then we want to match two 3d pictures that come from the same environment with a different shooting angle and get the transformation, which is the main task of our work.
The process is divided into five steps: preprocessing the point clouds, keypoint extraction, computing feature descriptors, coarse registration (Sampling and Consensus – Initial Alignment) and fine registration (Iterative Closest Point). All processing is done using the PCL library in C++.
2.1 Point Cloud Data Generation:
In order to test the performance of point cloud registration with different descriptors on point cloud data with different degrees of noise. We used the RGB-Depth 3D camera in Unity to capture 12 sets of dense point cloud data with varying degrees of noise in a flat scene.
Input point clouds are pre-processed as needed: filtering to remove outliers, down-sampling, removing NaN values.
2.3 Keypoint Detection:
In this step, one of the two methods are used : ISS (Intrinsic Shape Signatures) and SIFT (Scale-invariant feature transform). Other methods are Harris3d, NARF
2.4 Computing Feature Descriptors:
In this step, five descriptors are evaluated: FPFH, SI, SHOT, CSHOT, SIFT (both feature extraction and descriptor), and also a new data-driven descriptor is explored theoretically: 3DMatch.
The first five descriptors are based on histogram and geometry of point clouds, while 3D Match is a deep learning technique to learn the most robust representation of key-points.
2.5 Coarse registration:
To get a good initial state for ICP, SAC-IA (Sample Consensus Initial Alignment) is used to register the point cloud before ICP. SAC-IA shares a similar idea with RANSAC (Random Sample Consensus), which is widely used in registration as well (with a known model to fit).
2.6 Fine registration:
After SAC-IA, we get the approximate transformation matrix which gives ICP a good initial state. ICP finds the correspondence between points in a greedy way (closest point). It starts with computing correspondences between two point clouds and computes and applies a transformation to minimise the distance between corresponding points. This process is repeated until convergence. For best performance, ICP should get point clouds which are not extremely misaligned, which is the reason SAC-IA is used before ICP. ICP can be based on two methods – SVD and Non-Linear Least Square.
3. Comparison and analysis of pipelines:
Implementation of pipelines:
(1) SIFT3D + RANSAC + ICP
(2) SIFT3D + Spin Image + RANSAC + ICP
(3) SIFT3D + FPFH + RANSAC + ICP
(4) SIFT3D + SHOT + RANSAC + ICP
(5) SIFT3D + CSHOT + RANSAC + ICP
Dependencies: Point Cloud Library , C++, Visual Studio 2019, Unity
Number of registered point pairs is used as a metric for performance. Figure 1 presents the number of registered points of each pipeline using different feature descriptors with increasing the noises of the point cloud from 0 (0%) to 4.0(100%) in our experiment. It can be clearly seen that first, SI descriptor has a relatively better performance with increasing noises. In contrast, SIFT descriptor that is the key points themselves detected by SIFT has the lowest performance. Moreover, when other geometry-based descriptors have lower performance, CSHOT descriptor that use color information have a better performance than others.
During the experiments, we also recorded the computational cost of these descriptors. Figure 2 presents the running time of calculating descriptors for source and target key points with increasing noises of the point cloud from 0 (0%) to 4.0(100%). We can clearly see that SI descriptor is the most efficient descriptor. In contrast, SHOT and CSHOT become the most computationally expensive descriptors under current radius. However, if we increase the searching radius, the efficiency of these descriptors will decrease by different degree.
Overall, SI performs the best among these descriptors. SI has a good balance between performance and efficiency. Particularly, SI is suitable for real-time applications.
However, the performance and efficiency of these descriptors highly depends on the parameters setting. For instance, during the experiments, we found that the efficiency of FPFH decreased dramatically with increasing searching radius. If we set a larger searching radius for the FPFH descriptor, it will become very slow, while other descriptors are not very sensitive to the increasing searching radius. In order to obtain a clearer and comprehensive understanding of these feature descriptors, we should carry out more experiments in tuning the parameters setting in the future.
Continue reading →
Gesture Recognition with LSTM Classification of sEmg Signals for Prosthetic Control
Developed by Irene Sorianini, Federico Gala
Supervised by Nicola Covre, Prof. Mariolino De Cecco and Prof. Davide Brunelli.
Conditioning System Development by Ing. Antonio Selmo
Supported by Mirolab, Università degli Studi di Trento
From zero to hand movement control, the Mirolab presents the ambitious thesis project developed by Engineering Master course students, Irene Sorianini and Federico Gala. The project implements a deep-learning application for gesture classification and control of the prosthetic hand by monitoring the forearm muscles. A sEMG acquisition prototype allows the monitoring of muscle activity with high performance, versatility and low production costs. The results returned a high level of accuracy during test classification, which was followed by an on-line test to assess repeatability.
A series of hand crafted dry electrodes applied to the forearm collects the sEMG signals which, after processing, are classified by a machine learning model that returns visual feedback to operate a simple robotic hand.
The high accuracy in classification is achieved through data augmentation methods that aims to restrict the unavoidable limitations rasing when dealing with human factors.
At the moment the system involved a LSTM machine learning model that has been compared with other deep neural network model, such as CNN and DNN, showing more robustness and accuracy. The mis-classification are minimized with the introduction of parallel LSTM neural networks dedicated to specific gesture recognition.
The acquisition system consist of five elements. A neoprene cuff with integrated electrodes, an analog conditioning system for system for signal filtering and amplification, a high-perfomance ADC version system and a raspberry unit for digital dataprocessing. The electrodes are obtained by cutting 16 simple metal plates of size 20x20x1 mm and soldering the copper connector. Wiring is via coaxial cables to reduce noise of the 50 Hz mains voltage. The conditioning system has an amplification gain of 500 while flexing the signal at 500 Hz with a low pass filter, 50 Hz with a notch filter and at 20 Hz with high-pass filter. The ADC samples the 8 channels at approximatly 32 KSPS (4000 Hz per channel).Continue reading →
Laboratorio Pandemico 1 Giugno 2021
Link a Video ed Appunti Lezioni 3 Giugno 2021
2. Statistica ed Incertezza 31 Marzo 2021
Link Lezione 24 Maggio 2021
This thesis proposes the use of image processing and data acquisition to estimate the pose of a subject.
An RGB camera (or a ToF camera) can be used for image acquisition.
The acquisition of joints accelerations through a network of wearable IMU sensors is managed with Raspberry and MQTT communication protocol is used to share the information.
A sensor fusion algorithm allows the union of information and the reduction of uncertainty.
Everything is used in virtual environment for the real-time movement of an avatar displayed with Hololens and a visual feedback of the movement.
Continue reading →