Advancements in computer vision have primarily concentrated on interpreting visual data, often overlooking the significance of contextual differences across various regions within images. In contrast, our research introduces a model for indoor scene recognition that pivots towards the ‘attention’ paradigm. This model views attention as a response to the stimulus image properties, suggesting that focus is ‘pulled’ towards the most visually salient zones within an image, as represented in a saliency map. Attention is directed towards these zones based on uninterpreted semantic features of the image, such as luminance contrast, color, shape, and edge orientation. This neurobiologically plausible and computationally tractable approach offers a more nuanced understanding of scenes by prioritizing zones solely based on their image properties. The proposed model enhances scene understanding through an in-depth analysis of the object context in images. Scene recognition is achieved by extracting features from selected regions of interest within individual image frames using patch-based object detection techniques, thus generating distinctive feature descriptors for the identified objects of interest. The resulting feature descriptors are then subjected to semantic embedding, which uses distributed representations to transform the sparse feature vectors into dense semantic vectors within a learned latent space. This enables subsequent classification tasks by machine learning models trained on embedded semantic representations. This model was evaluated on three image datasets: UIUC Sports-8, PASCAL VOC - Visual Object Classes, and a proprietary image set created by the authors. Compared to state-of-the-art methods, this paper presents a more robust approach to the abstraction and generalization of interior scenes. This approach has demonstrated superior accuracy with our novel model over existing models. Consequently, this has led to an improvement in the classification of scenes in the selected indoor environments. Our code is published here: https://github.com/sebastianlop8/Semantic-Scene-Object-Context-Analysis.git

This article addresses the task of classifying scenes in typical indoor environments navigated by robots, using pre-trained convolutional neural network - CNN models with ImageNET and PLACES 365 datasets. The implemented models are the CNN VGG16 and RESNET50 architectures, which underwent various manipulations such as freezing and modification of intermediate layers. The performance of these networks in scene recognition tasks was analyzed, where a “scene” refers to an image that captures a partial or complete view of an indoor environment, displaying objects and their spatial relationships, and is similar to the scenes that mobile robots must perceive in their navigation task. To achieve this objective, a set of custom images JUNIO20V1, FEBR20V3, JUNIO20V3 and others, as well as 2 state-of-the-art image sets 15-Scenes and Sports, were subjected to various manipulations such as rotations, perspectives, sectioning, and lighting changes, creating a diverse image bank for both training and testing, allowing the analysis of the performance of the different modified CNN structures. The obtained results demonstrate the strengths of the different CNN structures in scene recognition tasks for application in indoor environments where mobile robots move.

The current article discusses the performance of local and global descriptors, as well as convolutional neural networks (CNNs), in tasks involving image recognition in interior spaces. The purpose of the test is to identify several realistic situations that closely resemble the typical working conditions for mobile robots. A robot interacting with its environment may be able to see portions of scenes in which objects are seen from various angles or changes in the lighting in various settings. The purpose is to investigate how well the different descriptors perform in identifying situations that meet the above criteria. In order to evaluate the effectiveness of visual descriptors and convolutional neural networks in the classification of images taken from the perspective of mobile robots in indoor environments, a proprietary database was implemented and subjected to several controlled transformations. These modifications made it possible to analyze the performance of Bag-of-Visual-Words (BoVW), Fisher Vectors (Fisher), Vector of Locally Aggregated Descriptors (VLAD), Global Image Descriptors (GIST), and CNN descriptors in visual categorization tasks according to the situational perception of mobile robots.The findings highlight the advantages of descriptors for the various test scenarios and highlight the need for hybrid models that employ both descriptors and CNNs for scene identification tasks in interior areas where mobile robots operate.

In recent years, the diffusion of the Colombian Coffee Cultural Landscape as the main tourist axis rich in ancestral traditions, is a country policy that seeks to promote the diffusion of heritage In recent years, the diffusion of the Colombian Coffee Cultural Landscape as the main tourist axis rich in ancestral traditions, is a country policy that seeks to promote the diffusion of heritage tourism and the conservation and promotion of the coffee heritage. Therefore, taking advantage of the growing advances in artificial intelligence (AI) applied to the tourism industry with emphasis on the cultural, artistic, historical and architectural diffusion with the different objects that characterize the culture of the Colombian Coffee Cultural Landscape. This article presents the application of CNN techniques focused on the detection and recognition of objects in the field of Colombian coffee cultural heritage, a line of research little explored. Although AI is just beginning to interact with the built environment through mobile devices, technologies in this field of object detection and classification have been producing and exploring digital models in different industrial sectors for a long time. The interaction between object detection algorithms and state-of-the-art information modeling is approached as an opportunity in heritage tourism as a central axis in a vision of making known the cultural, artistic, architectural and archaeological richness of this area of the country.

This article describes the implementation of a fuzzy and a PID control system for the velocity and torque of a DC power motor, based in the variation of the armor current and its duty cycle. The control system has been applied to a medium power DC motor, using Python, and a low-cost embedded system - Raspberry Pi, this work is fundamental given the importance of implementing velocity and torque controllers that yield energy optimization and correct operation being achieved through the regulation of the duty cycle applied to the power of the DC motor.

Contexto: La robotica movil continua siendo un area de constante actualizacion, donde se busca tener aplicaciones que permitan mejorar la calidad de vida de los seres humanos. Con base en lo anterior, en este articulo se presenta un novedoso diseno de un controlador de posicion y movimiento para un robot movil diferencial, utilizando redes neuronales artificiales y cuyo fin es llevarlo a aplicaciones reales en diferentes campos de accion. Metodo: La propuesta presentada esta basada en un controlador PID, el cual ha sido sintonizado con una red neuronal y requiere conocer el modelo cinematico del robot. Resultados: Las simulaciones muestran la eficiencia del control de posicion para el seguimiento de caminos explicitos, en este caso para el seguimiento de trayectorias rectilineas y curvilineas. La interfaz grafica presentada permite ajustar las ganancias con facilidad y verificar el seguimiento de la trayectoria en linea. Conclusiones: De los resultados obtenidos se puede concluir que la estrategia propuesta es de facil implementacion ya que se requiere informacion que de manera general entregan los sensores de un robot movil. Finalmente la sintonizacion realizada del controlador PID con la red neuronal permite obtener un correcto desempeno para seguimiento de trayectorias.

This paper presents the implementation of a differential mobile robot, using reconfigurable hardware (FPGAS-Spartan 3E) for navigation in dynamic environments. Different algorithms were implemented in VHDL, for navigation tasks, such as controlling the speed of the robot motors; reading the infrared sensor module (which was designed and implemented), for detecting obstacles in the navigation environment; digital compass reading to determine the robot’s direction, and a Matlab GUI for exchanging information between the robot and the controller. The algorithms and the results of some tests conducted with the robot in different environments are shown.

Context: Mobile robotics remains being an area of constant updating, which seeks to have applications to improve the life quality of human beings. Therefore, the paper presented a novel design of a position and move controller to a differential mobile robot using artificial neural networks, which aims is to bring a real applications in different fields. Method: The proposal is based on a PID controller, which has been tuned using a neural network and requires knowledge of the kinematic model of the robot. Results: The simulations show the efficiency of the control position for following paths, in this case for tracking straight and curved paths. The graphical interface allows adjusting the gains easily and verifies the trajectory tracking online. Conclusions: From the results we can conclude that the proposed strategy is easy to implement since the required information is generally deliver by mobile robot sensors. Finally, tuning PID controller with the neural network allows obtaining a correct performance for trajectory tracking.

Este articulo presenta varias tecnicas para el reconocimiento de diferentes tipos de superficies muy comunes en ambientes de navegacion de robots moviles. El reconocimiento de estas zonas se inicia en la caracterizacion de las senales de un sensor de rotacion de ultrasonido. Las areas que se han estudiado son bordes, esquinas, paredes y vacio. Para realizar la caracterizacion, las senales se procesan utilizando diferentes familias de la trasformada wavelet para su posterior analisis y clasificacion utilizando un clasificador bayesiano. Finalmente, se lleva a cabo una comparacion de las diferentes familias de wavelets utilizadas para determinar la mejor de ellas.

Ultrasonic sensing is a well suited cost-benefit technique used in mobile robotics. Nevertheless, this sensor type has a key drawback: its very Wide beam. This fact causes that this sensor becomes useless in the environment map building process. Another problem attached with this sensor is that it offers very little facilities for landmark extraction and object recognition. This problem is known as Matching. In this paper we introduce an ultrasonic sensor capable to extract wall or corner from an indoor environment, as well as to reduce the angle uncertainty, on the based of echo amplitude reaching the transducer. Moreover, we developed the stochastic model for this sensor as well as a matching algorithm for using this device as main observation element for probabilistic SLAM (Simultaneous Localization And Mapping).

The use of mobile robots is justified in applicatio ns where the tasks present risk to human life. The transport of dangerous materials , mineral excavations, or the inspection of nuclear plants are examples where a m obile robot can perform its work. To realize these tasks is necessary to equip the robot from a set of sensors that give it the ability to navigate in any environment. This paper presents the design and implementation of a sensor package, whic h will provide at the mobile platform of the intelligence needed to navig ate in any environment.

The present article describes a block of sensorial perception, using infrared sensors, which was implemented in an FPGA's, Sparta n 3E of Xilinx company.

These papers present the implementations of measurements system of distance with infrared sensors and programmable logical dispositive for application of building environments maps. Shows obtained results from infrared sensors and utilization of differents technological basis microcontroller and FPGAS, working shincronized

En este articulo se implementa un sistema de control para la navegacion reactiva de un robot movil en ambientes dinamicos, basado en comportamientos difusos. Un anillo de sensores ultrasonicos es utilizado para detectar los obstaculos del entorno por donde se desplaza el robot. El controlador difuso se compone de cuatro comportamientos basicos y su respectivo conjunto de reglas, dichos comportamientos son: navegar por pasillo, seguir pared, alcanzar objetivo y evitar obstaculos. Los algoritmos implementados para cada comportamiento utilizan funciones de pertenencia del tipo triangular. El control desarrollado se simulo e implemento en la plataforma movil P-METIN, construida por el grupo GIROPS. Los resultados obtenidos son satisfactorios para navegacion en ambientes dinamicos.