Jose Reginaldo Hughes Carvalho

Genetic Algorithm Approach for a Class of Multi-Criteria, Multi-Vehicle Planner of UAVs

This work presents an solution for the problem of planning the routes of a small fleet of hand-launched fixed-wing UAVs to monitor by vision an area of forest at the minimum possible time, for rescuing support, fire detection, deforestation mitigation, among other important applications in the Amazon rain forest operational scenario. However, time is not the only criteria to be considered. Given that it is virtually impossible to recover a missing airplane in the deep jungle, another relevant objective is to apply as fewer UAVs as possible, to reduce the risks for the whole UAV system. Moreover, another aspect that complicates this specific application is the non-holonomic constraint of the fixed wing airplanes (lack of instantaneous lateral velocity). We had to restrict the changing of heading, and thus, the roll angle, to prevent the drifting of the PoI (point-of-interest) out of the camera plane. As a consequence, long turn radius imply longer route. The authors modeled this problem using the multi vehicle routing problem, combined with Dubins path generator to address the non-holonomicity aspect, in a multicriteria formulation to reach a solution that simultaneously takes into account time, number of UAVs and expended resources. The resulting problem formulation has a computational effort that tends to be extremely high as the number of way-points increases. The authors applied Genetic Algorithm to solve this multi-criteria NP-hard Dubins adapted routing problem. The presented results, applied to a fleet of mini-UAVs, show the effectiveness of the proposed work in providing a satisfactory solution in a very reasonable execution time.

Generation of critical mass in education: An initiative to engagement

The lack of professionals is considered a major risk to the Industrial Pole installed in Manaus, Brazil. Specially, STEM education is intricately difficult and if the student is not well involved, supported and motivated, the chances of success are very low, which can be verified by the 25% to 30% dropout rate per semester in our institution. Given this scenario, a qualification program denominated “multiple vortexes of know-how” was conceived to address three aspects: (i) reduce student dropout; (ii) enlarge the reachable community; (iii) offer different levels of knowledge. Vortex is in essence an action, working interconnected and coordinated to other vortexes. Four different actions were prepared: classroom disciplines; a talent development program; holding short-term events, and enrolling in intensive training. This present work describes the vortex that enrolled the major number of students: the classroom disciplines. A catalog of optional undergraduate and graduate disciplines was offered to the students. The best students received a prize based on their performance. In summary, we had around 650 students qualified with 84 scholarship prizes granted. The paper will present as a contribution to the program, the catalog of disciplines and recommendations on how to proceed with the activities.

Enhancement of Degraded Images by Natural Phenomena

The efficiency of environmental monitoring through imagery data is strongly dependent on the quality of the acquired information, despite weather conditions or other uncontrolled degradation factor. This article describes a series of combined techniques of image enhancement to partially recover information “lost” due to unfavorable operational conditions or natural phenomena, such as: fog, rainstorms, underwater dust (green dust), poor illumination, etc. We based our approach on a process known as homomorphic filtering, which is intrinsically related to the transformation from the spatial to the frequency domains, directly involving the Fourier Transforms, followed by specific enhancement techniques, such as Clipping and Stretching. Although, the use of these techniques separately, without the proper adaptation and coupling, can result in damaging even more the image, the authors developed an efficient sequence of enhanced filtering able to recover most of the affected information. Moreover, the proposed methodology proved to be generally applicable to a large class of images in poor conditions, with a performance comparable to the methodology

Feature Transform Technique for Combining Landmark Detection and Tracking of Visual Information of Large Rain Forest Areas

Researchers have been spending a lot of effort in increasing the level of autonomy of Unmanned Aerial Systems (UASs). There is a sort of important scenarios where an autonomous drone would be very effective. One of these scenarios of applications is the long term monitoring of the Amazon rain forest. The uniform pattern of the canopy defines a mission difficult to be performed by a human operator. Imagine someone in front of a monitor seeing for hours long the very same thing: treetops. In such situation, an embedded vision system capable to drive the vehicle while taking decision of what is not fitting to a standard canopy pattern plays a critical role on both remotely operated and autonomous navigation modes. The goal of this work is to present a scheme based on image processing able to extract natural landmarks in forest areas, and to track them during posterior missions over the same area, as reference for the onboard navigation system. The scheme is composed of two main steps: 1) Nonrelevant features suppression based on wavelet, to eliminate the canopy uniform pattern, and 2) Key points extraction by SIFT algorithm, to extract new landmarks or to track existing ones. Preliminary results demonstrated that this system can increase the robustness of mission execution in scenarios where usually only GPS references are available.

Higher Education in the Amazon: Challenges and Initiatives

In this paper, we present an overview of the results achieved by a few applications of educational research towards overcoming the difficulties in the Amazon region’s educational scenarios for a better higher education. Manaus, the capital of the Brazilian state of Amazonas, demands an urgent increase in the number of qualified professionals, mostly for its Industrial Pole. However, Amazonas has a poor record of educational performance in all scholastic grades. The situation is especially critical in the rural areas, where the population, separated by miles of uncharted jungle, is usually only accessible through the rivers. This geographic isolation transforms the simplest educational activities into a daily challenge. A sustainable solution to the lack of professionals needs the ability to reach the students with learning activities wherever they are. In this work we share experiences in two different scenarios: the Capital and the rural areas. For the capital, the authors began a large-scale program to address three aspects: (i) reducing student dropouts; (ii) enlarging the reachable community; (iii) offering different levels of knowledge. Four different learning activities were prepared: undergraduate and graduate classes, a talent development program, short-term events and intensive training. For the rural towns, the authors describe the educational experience with undergraduate and graduate distance learning and the ongoing research using learning analytics and adaptive models based on the LMS Moodle. The results are encouraging, as the measurement of the positive impacts of the reported initiative clearly indicates that a structured program would bring solid and long lasting benefits. It is also worth noticing that the proposed solutions may also fit other learning scenarios.

A Comparison on Supervised Machine Learning Classification Techniques for Semantic Segmentation of Aerial Images of Rain Forest Regions

Segmentation is one of the most important operations in Computer Vision. Partition of the image in several domain-independent components is important in several practical machine learning solutions involving visual data. In the specific problem of finding anomalies in aerial images of forest regions, this can be specially important, as a multilevel classification solution can demand that each type of terrain and other components of the image are inspected by different classification algorithms or parameters. This work compares several common classification algorithms and assess their reliability on segmenting aerial images of rain forest regions as a first step into a multi-level classification solution. Finally, we draw conclusions based on the experiments using real images from a publicly available dataset, comparing the results of those classification algorithms for segmenting this kind of images.

A Visual Data Fusion to Aerial and Land Robots Odometry

This paper aims the estimation of the robots motion angles through a linear fusion of two visuals information: Horizon Line position and orientation, and Image Optical Flow. The need of integration of movement estimation techniques is due to the fact that the angles estimation using Horizon information is used with efficiency in several works about aerial robots odometry, but this approach does not enable yaw angle estimation. Optical Flow approach enables the estimation of all of the angles and can be used without the presence of Horizon on the vision field, but can fail in situations which its constraints are not complied. Preliminary experiments made with aerial robots proved the efficiency of the presented method. Despite the difference on the degrees of freedom of other types of robots in comparison with aerial robots, this odometry method can be used on all the types of land robots to estimate its movements. Obviously, in particular cases, the land robot movement can be described with less of the three movement angles because the minor degree of freedom of the robot.

Natural Landmark Tracking Method to Support UAV Navigation over Rain Forest Areas

Field application of unmanned aerials systems (UASs) have increased in the last decade, with special focus on missions in which there is any kind of risk for the crew. However, in some situations the lack of onboard pilots makes the mission execution a very difficult task. As an example one has long endurance missions over rain forest, and glacial areas, due to the uniform patterns. In this scenario an embedded vision system plays a critical role on both remotely operated and autonomous navigation modes. This paper presents an approach towards a vision system model able to track landmarks in forest areas, as a navigation support for UASs. This system consists of two main steps: (1) multiresolution decomposition and nonrelevant features suppression techniques based on wavelet; and (2) automatic control points identification and matching for sensed video frame and reference image. Preliminary results demonstrated that this model can correctly track a sequence of natural landmarks while suppressing nonrelevant features.