9616 research outputs found
Sort by
Recognition and Sorting of Small Objects Using a Robot and Computer Vision
No full textRoboti so danes nepogrešljiv del sodobne industrije, saj zaradi svoje natančnosti in ponovljivosti odlično opravljajo ponavljajoča se opravila. Z razvojem in integracijo strojnega vida ter umetne inteligence so postali bolj prilagodljivi ter primerni za kompleksnejše naloge, kot je na primer razvrščanje objektov. Pri takih opravilih pogosto presegajo človeške sposobnosti. Namen magistrskega dela je razvoj in implementacija sistema, ki z uporabo kooperativnega robota in strojnega vida samostojno prepoznava ter razvršča manjše objekte. Za kooperativni robot UR3e smo izračunali Denavit-Hartenberg parametre ter napisali programsko kodo za izračun transformacijske matrike med njegovo bazo in vrhom za poljubne kote v sklepih. Za prepoznavanje štirih kategorij manjših objektov smo v programskem jeziku Python razvili program, pri čemer smo naučili nevronsko mrežo na osnovi arhitekture ResNet18, nato pa program za prepoznavo implementirali na napravi NVIDIA Jetson Orin Nano. Ločeno smo izdelali program za prijemanje in premikanje manjših objektov z uporabo robota Universal Robots UR3e. Mini-HPC NVIDIA Jetson Orin Nano in robot Universal Robots UR3e smo medsebojno povezali preko Modbus komunikacije in tako razvili celovit sistem.Robots have become an indispensable part of modern industry, as their precision and repeatability make them ideal for performing repetitive tasks. With the integration of computer vision and artificial intelligence, robots are becoming capable of handling complex tasks such as object sorting, often surpassing human performance. The purpose of this master’s thesis is the development and implementation of a system that combines a collaborative robot with computer vision to autonomously recognize and sort small objects. For the UR3e cobot, the Denavit–Hartenberg parameters were calculated, and MATLAB code was written to compute the transformation matrix between the base and the end-effector for arbitrary joint angles. To recognize four categories of small objects, a program in Python was developed, training a neural network based on the ResNet18 architecture and deploying the recognition system on an NVIDIA Jetson Orin Nano. In addition, a separate program for manipulating small objects was created for the UR3e robot. Communication between the Jetson Orin Nano and the robot was established using Modbus, resulting in a fully integrated system
Biotransformation of xylan isolated from sunflower meal into xylo-oligosaccharides using endo-xylanase immobilized on oxidized nanocellulose
No full textIn recent years, oxidized nanocellulose (Ox-CNC), obtained via periodate oxidation, has been widely used in foods, cosmetics, and biomedical fields due to its unique characteristics such as biodegradability, antimicrobialactivity, non-toxicity, crystalline structure, and presence of reactive dialdehyde groups on its surface. Herein, a novel sustainable, cost-effective, and clean approach using xylanases enzyme preparation ROHALASE® SEP-VISCO immobilized on Ox-CNC was applied and utilized for the transformation of xylan, isolated from agro-food waste, into xylo-oligosaccharides (XOS), considered as emerging rebiotics. The study has confirmed the efficient covalent immobilization of xylanase onto the obtained dialdehyde-modified cellulose nanocrystals (Ox-CNC) with the yield of xylanase immobilization of 55 %. The highest specific activity of the immobilized xylanase (1.5 IU/mg proteins) was achieved in optimal conditions: initial enzyme concentration of 400 mg/g of support, pH of 9.5, and a reaction time of 3 h. This study proved concept of the application of cellulose nanomaterials in green processes for the production of potential prebiotics from agro-food waste since immobilized enzyme was applied in reaction with xylan obtained by fractionation of sunflower meal (SFM). Ox-CNC-xylanase demon- strated an XOS productivity yield of 31 % based on the total SFM xylan content after 3 h. Additionally, the
prebiotic activity was confirmed in a microbiological test within which the resulting mixture rich in XOS selectively stimulated the growth of some commensal gut microbes (Lactiplantbacillus plantarum 299v, Lactica-seibacillus rhamnosus GG and Saccharomyces boulardii CBS 5926) in comparison with pathogenic species (Escherichia coli ATT 25,922)
How to extract hydrogen bonding energies from circular dichroism measurements of polypeptide solutions?
No full textThe temperature dependence of helicity degree is a final point for
extracting data about the conformational transitions in proteins from the CD
signal [1]. But it is also the starting point for extracting, processing and
analyzing the properties of a particular protein [2]. All one needs is to fit
experimental data to a proper model which accounts for the most relevant
interactions in the system. In particular, water-protein and intra-protein
hydrogen bonds dominate the behavior of proteins and a theoretical model
describing such interactions is of vital importance. In my talk I will present a
model of this kind, we have developed recently. We start from the microscopic
Hamiltonian formulation of the polypeptide model in water and follow the
usual Statistical Mechanics route from the model to partition function and to
the thermodynamics. Using Mayer expansion and summation over the solvent
degrees of freedom, the problem is shown to be equivalent to in vacuo model
with some effective, temperature dependent interaction energy [3]. Estimated
partition function leads to the expressions for the thermodynamic potentials and
order parameter averages. The comparison (least-square fit) with the
experimental data points from CD experiments on protein folding allows to
extract the information on hydrogen bonding strengths, not available with the
classical approach [4,5]. We have developed a free online tool [6], FitFoldData
(https://fit-fold-data.ung.si/cd-fitting), which performs the above-mentioned
procedur
Polysaccharide degradation in an Antarctic bacterium
No full textGlycoside hydrolases (GHs) are enzymes involved in the degradation of oligosaccharides and polysaccharides. The sequence space of GHs is rapidly expanding due to the increasing number of available sequences. This expansion paves the way for the discovery of novel enzymes with peculiar structural and functional properties. This work is focused on two GHs, Ps_GH5 and Ps_GH50, from the genome of the Antarctic bacterium Pseudomonas sp. ef1. These enzymes are in an unexplored region of the sequence space of their respective GH families, not allowing a reliable sequence-based function prediction. For this reason, a computational pipeline was developed that combines deep learning “dynamic docking” on AlphaFold 3D models with physics-based molecular dynamics simulations to infer their substrate specificity. From in silico screening of a repertoire of potential oligosaccharides, only xylooligosaccharides for Ps_GH5 and galactooligosaccharides for Ps_GH50 emerged as catalytically competent substrates. Biochemical characterization agrees with computational simulations indicating that Ps_GH5 is an endo-β-xylanase, and Ps_GH50 is active mainly on small galactooligosaccharides. In conclusion, this study identifies two novel GHs subfamilies placed in remote regions of the sequence space and highlights the efficacy of substrate specificity prediction by computational approaches in the discovery of new enzyme