Abstract The quality assessment of fruits and vegetables is crucial in the agroalimentary supply chain, as it directly affects consumer satisfaction, market value and overall food security. Traditional approaches rely on visual inspections or destructive techniques, which are labor-intensive and time-consuming. On the contrary, non-destructive techniques emerged as promising alternatives, offering solutions that can be adopted in real environments. Previous studies emphasized that the color distribution over images plays a significant role in the quality evaluation of food. In this paper, we propose a solution that leverages an autoencoder architecture to extract groups of relevant colors from the complete histogram of colors. To enhance the analysis of real-world images with complex backgrounds, we employ a pre-trained U2-Net architecture for background removal. Moreover, we propose a novel procedure based on outlier detection to identify and remove parts of the background that are not fully eliminated, especially along the edges of the product. After this preprocessing, we extract a complete color histogram which is fed to an autoencoder architecture, to extract high-level features representing color groups at different levels of granularity. The goal is to make the learned models less sensitive to light and color perturbations. Our experiments, conducted on two real-world datasets related to two different learning tasks, demonstrated the effectiveness of the proposed solution, that outperformed several baseline and state-of-the-art approaches, also based on complex neural network architectures.

Service robots are expected to be used in many household in the near future, provided that proper interfaces are developed for the human robot interaction. Gesture recognition has been recognized as a natural way for the communication especially for elder or impaired people. With the developments of new technologies and the large availability of inexpensive depth sensors, real time gesture recognition has been faced by using depth information and avoiding the limitations due to complex background and lighting situations. In this paper the Kinect Depth Camera, and the OpenNI framework have been used to obtain real time tracking of human skeleton. Then, robust and significant features have been selected to get rid of unrelated features and decrease the computational costs. These features are fed to a set of Neural Network Classifiers that recognize ten different gestures. Several experiments demonstrate that the proposed method works effectively. Real time tests prove the robustness of the method for realization of human robot interfaces.

Distributed networks of sensors have been recognized to be a powerful tool for developing fully automated systems that monitor environments and human activities. Nevertheless, problems such as active control of heterogeneous sensors for high-level scene interpretation and mission execution are open. This paper presents the authors' ongoing research about design and implementation of a distributed heterogeneous sensor network that includes static cameras and multi-sensor mobile robots. The system is intended to provide robot-assisted monitoring and surveillance of large environments. The proposed solution exploits a distributed control architecture to enable the network to autonomously accomplish general-purpose and complex monitoring tasks. The nodes can both act with some degree of autonomy and cooperate with each other. The paper describes the concepts underlying the designed system architecture and presents the results obtained working on its components, including some simulations performed in a realistic scenario to validate the distributed target tracking algorithm.

This paper describes a technological solution for the remote fruition of an archaeological site, a cave hosting pre-historical heritage of relevant importance, using a properly equipped mobile robot. The goal is twofold: to increase the fruition of these precious paintings (both off-line and on-line with the robot moving in the cave) and to collect data (about the two-dimensional map of the complex cave and about the three-dimensional structure of the wall hosting the paints) that can be important to increase the understanding of the environment by the experts. The choice of the robot is suitable for this task before it does need a very light and minimal installation of mobile and reversible infrastructures: in fact in such an environment any intervention can cause significant damages to the archaeological site. The current modules are able to explore the cave to determine the most interesting zones (with a richer content of pre-historical wall paintings), to collect three-dimensional data (using a laser range finder) that allow the reconstruction of a planar map of the cave, the navigation between these relevant zones, the fine positioning with respect to the paintings and the reconstruction of a full three-dimensional model of the walls hosting the paints starting from a set of three images acquired from slightly different points of view.

The paper presents a fast and reliable approach to estimate body postures in outdoor visual surveillance. It works on patches corresponding to people, recognized by two subsystems (motion detection and object recognition) on image sequences coming from a still camera. The proposed algorithm is based on an unsupervised clustering approach and is substantially independent from a-priori assumption about the possible output postures. Horizontal and vertical histograms of the binary shapes associated to humans are selected as features. The Manhattan distance is used for building clusters and for run-time classification. After experimental tests the BCLS (Basic Competitive Learning Scheme) algorithm has been selected for the construction of clusters. The whole approach has been verified on real sequences acquired while typical illegal activities involved in stealing were simulated in an archeological site.

The main aim of this work is to present a simple and reliable approach to the estimation of human body postures. The applicative context is the visual surveillance of an archaeological site. Motion detection and object recognition subsystems process image sequences coming from a still camera. Whenever a human is detected, his postures are characterized by the proposed pose estimation module. Then the results are fed to a HMM subsystem that identifies the current activity of the examined subject The proposed algorithm is based on an unsupervised clustering approach that makes the system substantially independent from any a-priori assumption about the possible output postures. The features selected for posture estimation are the horizontal and vertical histograms of binary shapes. A modified version of the Manhattan distance is used for both cluster identification and for run-time classification. After extensive experimental tests with different clustering schema, the BCLS algorithm (basic competitive learning scheme) has been selected. The proposed approach makes possible to change the number of classes, during the classification phase, without repeating the training step. Moreover it provides a measure of the reliability of its results. The proposed method has been verified on sequences acquired while typical illegal activities involved in stealing were simulated in a real archaeological site.

In this paper we describe a people tracking system addressing the visual surveillance of outdoor environments. This context involves the detection of objects moving in a sequence of images acquired by a TV camera and the recognition and tracking of human figures to allow the analysis of their gestures. The proposed tracking approach does not consider the human body parts as features to be tracked, nor uses a priori body models. Moreover, also classical prediction models are useless in the specific context (visual surveillance of archaeological sites) because the behaviors identified as illegal imply a sequence of unpredictable gestures performed in a small area. The proposed method is based on the estimation of similarity between human figures segmented in the temporal sequence. The similarity score is estimated as a function of the number of corresponding feature points. The system has been verified on real image sequences acquired by a static TV camera while the gestures normally performed by intruders were simulated in a real archaeological site.

Our application context is the visual surveillance of archeological sites. In this context the main aim is to detect the presence of people and to scan them in order to recognize intruders on the basis of their gestures. Since an intruder needs some utensils indispensable to perform the illegal actions of excavating on the ancient ruins, intruder detection involves first of all to ascertain if a person is carrying some objects and then recognizing the kind. In this paper we concentrate on the recognition of the objects carried by the detected moving persons. An example-based learning technique is used to first detect people and successively to scan them to recognize the possible carried objects. The patterns to be analysed are represented through the approximation coefficients of their three level wavelet decomposition. Pattern classification is performed through a supervised three layer neural network.

Realizing complex robotic tasks by combining elementary behaviors is simple, robust and effective. Learning controllers from examples avoids manual identification of system states and actions and direct definition of their relationships. Fuzzy controllers represent direct associations between input and output variables, typical of reactive behaviors, in a form understandable by both automatic tools and humans. The paper describes the application of a set of tools developed for supporting the automatic design of fuzzy controllers (recording of associations between control commands and system states, identification of the required controller's components). Safe navigation of a real vehicle in an unknown indoor environment is addressed by combining two behaviors, wall-following and door-crossing, using a third module for judging their competence in each situation. Its performance, driving a Scout in our offices, is shown and related to the limitations of the two basic controllers.

We describe a navigation approach, based on a reinforcement learning algorithm, that allows a mobile robot to move in an unknown indoor environment learning autonomously in a few trials the actions for reaching a particular goal location. The control architecture merges visual information and sonar readings to evaluate the state of the system to which the learning algorithm relates the best movement for reaching the goal. As a result, after limited experience, the robot learns efficient behavioral sequences and improves its learning incrementally. In addition the learning system adapts well to new situations of the environment or new task requirements. The results obtained in simulation and in real experiments, carried out in our laboratory, have shown that our system is tolerant to noise in sensor measurements and during each trial is always able to reach the goal.

This paper deals with oriented texture analysis applied to a problem of defect detection and classification in leather inspection, for industrial application. In this paper we present results obtained using a defects detector based on oriented texture analysis, which reveals itself useful for a few classes of leather defects, such as scars or folds. These defects can be detected by using a black and white camera running over the leather patch and by classifying textures, on the basis of their gradient orientations and local coherence. Leather defects are detected by segmenting the oriented texture map of the leather surface. The orientation in every point is expressed using a normalized B-spline basis whose knots are distributed on the leather surface. The resulting erector of coefficients can be used to perform the final segmentation and localize defective areas. A morphological segmentation procedure is applied to the regularized oriented texture field in order to extract probable defective areas.

This paper deals with the problem of recovering directionality from compositional textures. It presents a new method that has proved to be effective for detection and classification of leather defects in the manufacturing industry. A flow field, estimated from texture, is represented as a set of projection coefficients onto suitable elementary basis vector fields. Since these bases are neither orthogonal nor complete, the coefficients must be estimated using a global optimization criterion. The method performs this optimization using a neural network. A region growing approach, based on single iterations of the same neural network, expands patches corresponding to oriented structures and produces the final segmentation map.

Fuzzy logic is a useful tool for realizing a direct mapping between perceptual situations and control commands in robotic applications that do not require internal representation or planning. It allows explicit programming and automatic learning from suitable training data to be mixed in several ways in order to produce the necessary control rules. Automatic learning enables both the reduction of the annoying and error-prone explicit programming work and the evolution in time of the controller in order to cope with dynamically changing environments. At any time an expert can verify and eventually modify the knowledge of the fuzzy system according to its personal experience. Two methods for the automatic extraction of rules from training data have been proved on a fuzzy control system for wall-following. The system is intended to grow including all the behaviors required for the safe navigation of a autonomous mobile vehicle and their arbitration. Distance measures supplied by an ultrasonic sensor ring have been chosen as sensory data. The training data have been collected during operator-driven runs of the vehicle. The same data have been used by the two methods for building rule bases that have proved effectively in driving a real AGV (a TRC Labmate base) in an indoor environment.

Presents an approach to detect moving obstacles in order to define a collision zone around them. The approach uses a camera system to provide image sequences from which the dynamic aspect of the scene can be recovered. In fact, if the camera samples the observation space at some time intervals, a two-dimensional image sequence may be created, due to the relative motion. The analysis of these sequences can be useful to recovery information about the relative notion between a mobile vehicle and the navigation space. The appearance of an obstacle can be recovered using optical flow analysis. The authors compute optical flow maps to solve the problem of detecting discontinuities in the flow field in order to segment the image in different moving objects. An estimation of depth values for the segmented points is accomplished by means of optical flow and focus of expansion. Experimental results on image sequences of a real scene are presented.>

Recovering three-dimensional visible surfaces is an important cue in computer vision. Like most of the inverse problems it is hard to be solved due to the insufficient and inaccurate data that can be collected using passive sensors. Data fusion and integration over time can be used to overcome this problem. In this paper the depth and orientation are acquired from a scene and are used together to build a dense map of the visible surface for each of several points of view. An incremental estimator is used to integrate these maps, as soon as they become available, in a more reliable result.>