Computer Vision and Object Recognition
A major challenge of computer vision is the recognition of occluded objects. This problem arises when only a portion of an object is visible. Recognition of partially occluded objects is an important task in many applications, such as target recognition and tracking, industrial inspection and robot navigation. These applications require robust recognition techniques that can handle a reasonable degree of occlusion.
A practical example of occluded object recognition is where a robot must recognize an industrial part in a composite scene of overlapping parts. A specific example of this is a scene comprised of overlapping industrial tools such as wrenches, hammers, strippers and pliers. Considering the problem of a robot picking up an industrial part from such a scene, the current work proposes a fuzzy model-based object recognition algorithm to allow the automatic recognition of partially occluded objects. The goal is to recognize the presence of tools from a database of known objects. The algorithm is intended to be rotation, translation and scale invariant.
Learning and Adaptation Using Fuzzy Inference
The human brain is extremely adaptable and robust when it comes to learning in novel environments. The goal of this work is to use an intelligent system, developed using fuzzy logic, to emulate human motor learning (Figure Below). Such a system may be useful for the development of learning mechanisms for applications in robotics and artificial intelligence, as well as in gaining an understanding of human learning to aid in rehabilitation.
The particular task under consideration is upper-limb reaching in the presence of an external force field. A simulated human arm attempts to move its hand along smooth straight-line paths between starting and target locations. Using the position and velocity errors from a given reaching movement, the fuzzy learning system employs a knowledge base of if-then rules to incrementally adjust a model of the experienced force field. The end result is that the actual path followed by the hand approaches the intended straight-line path.
Vision-Based Robotic Arm Manipulator
Abel Bustillos, Gilberto Contreras, Rod, Quinn, Adrian Trejo
Robotics has a wide variety of applications in most major production facilities around the world. For example, the electronics industry uses robotic manipulation for the assembly of sensitive components, such as microprocessors. The materials processing industry also makes use of robotics to safely process both corrosive and non-corrosive materials. However, the use of robotics is best known in the automotive industry. In this case, auto makers employ a continuous robotic automation process to increase the production of automobile manufacturing as well as increase sales. Thus, robotics remains an integral part of any industry and further research remains essential in the Laboratory for Industrial Metrology and Automation.
One project that is currently under design is the development of a Vision-Based Robotic Arm Manipulator. Such a manipulator is an automated arm that uses a camera-based sensor to detect and retrieve a specific object. The robotic arm manipulator then transfers the object to a designated area and repeats the process at the discretion of the operator. Additionally, a VBRAM can also be programmed to distinguish and retrieve different objects. Furthermore, improvements in operation efficiency and robotic design may also be implemented at any time to extend the robotic project for future engineering students. Consequently, this research project not only accelerates interest in robotic innovation, but also prepares engineering students for multidisciplinary responsibilities.