Estimating surface depth from two digital images taken from the same viewpoint

New Scientist reports on new computer vision research that allows the learning of surface structure using two digital images taken from the same viewpoint but under different illumination. More specifically, one photo is taken with flash and another using only ambient light. The differences between the two images allowed researchers from the University of Manchester in the UK and Dolby Canada to extract albedo and depth information for the surface; such information is then used to render the surface from different viewpoints properly illuminated. Under the title of “surface depth hallucination,” the researchers presented their work at the recently held, well known and very prestigious SIGGRAPH conference.

The project's web site describes the method as follows.

Surface depth hallucination offers a simple fast way to acquire albedo and depth for textured surfaces that exhibit mostly Lambertian reflectance. We obtain depth estimates entirely in image space, and from a single view so there are no complications that arise from registering texture with the depth obtained. (source)

The following video from New Scientist explains the surface depth hallucination method in more detail and it includes examples of estimated surface models.

The Toyota Winglet

About a month ago Toyota Motor revealed a new stand-up scooter (called the Winglet) that it was creating. Since the announcement there has been plenty of buzz around the internet and now the Winglet is seen as competition for the Segway in the personal transport market.

The Winglet consists of two wheels, an electric motor, a body, and internal sensors that monitor the rider’s position. The sensors are able to detect stability issues and quickly resolve them by making power adjustments, hence keeping the Winglet (and the rider) from falling. The technology is quite advanced, but like the Segway, the market isn’t very big for it at the moment. Perhaps Toyota and Dean Kaman (developer of the Segway) are ahead of our time and one day their creations will be widely used.
Even if the personal transport business never kicks off, it is still amusing to look at the way these things work. The Winglet constantly detects the speed and position of itself in order to keep the rider standing. It can go up to 3.7 mph, which is slow in comparison to the Segway (which hits 12.5 mph), but is still nice for casual movement.

Toyota has created the transport in several sizes to accommodate people of different sizes. The height of the Winglet can range from 18 inches all the way to almost four feet tall. The handlebars can also be adjusted. The width and length is standard in all sizes at 18 and 10 inches respectably.

The presentation only displayed prototypes of the creation, but Toyota will be developing more soon. The company plans to start testing the Winglets this fall in Japan and they will move forward from there.

Tongue computing technology may help disabled

Researchers at Georgia Tech believe that the mouth can essentially be turned into a computer via a tiny magnetic system placed on the tongue. The tongue would be the controller and a user would touch different teeth to tell the system what to do. Different sensors on the cheeks send the information to a head unit that calculates the signal and performs the requested task. The technology has drawn a lot of interest because it would be very useful for many disabled people.

The Tongue Drive System would allow handicapped users to move around their wheelchairs, control computers or other electronic devices, and manage home appliances with simple movements of the tongue. Mike Jones, vice president of research and technology at the Shepard Center in Atlanta, said that the device would give almost a limitless number of “switches” to do different things.

It is the huge number of controls that really makes this technology intriguing. There are currently several other devices that allow the disabled to control thing with their face and mouth, but this is the first that allows so many options with such a simple approach. Currently, the most-used systems of this type only have a few commands, which really don’t allow the user to do many things. There are also systems that have hundreds of different controlling options, but at the moment those are very complex and often difficult to use and set up. The Tongue Drive System brings together the best of both worlds with its simplicity and maneuverability.

British Army urban reconnaissance contest

Last week the British Army held a contest in an effort to spark the development of new technology for urban reconnaissance. The event had a great turnout with devices and gizmos of all sorts being presented including various types of robots, flying saucers, and remote-controlled buggies. Some of the entries were very intricate and advanced, while others weren’t.

The winners were announced on Tuesday and Stellar Research Services from Southampton came out on top. Their creation was a combination of a flying robotic aircraft, ground drone and a tracked vehicle. Reporters have compared the robot to WALL-E, Pixar’s robot character, due to similarities in appearance.

Stellar’s creation, called Saturn, scored the most points in the complex obstacle course set up by the contest holders. The robot had to explore a mock village while looking out for and detecting threats such as hidden snipers, armed enemies in the open and roadside bombs. Saturn was the best of the competing robots at identifying and locating the targets, hence getting the team the most points. The three-in-one robot used a wide array of sensors, cameras, and radar in searching for the “threats.”

Team Leader Julia Richardson was extremely pleased after her team won, adding that they had worked long and hard to make it happen. Several other teams were also handed accolades for unique and inspiring designs. The army got a good turnout and is likely to begin adapting some of the ideas from the contest into actual utilities for use in the battlefield.

UK company develops snake-arm robot

England-based OC Robotics has created a snake-arm robot which is very flexible and also fully controllable. The United States Department of Defense (DOD) will be testing the device in small and unusual areas in order to see if it will be of use to them. The creators claim that it can be controlled very precisely and could prove to be very useful for investigations.

The snake-arm is 24 inches long and consists of a “backbone” which is broken into segments by little wires. This allows a user to control each of the wired segments via joystick, meaning it can be bent in several locations so that it can be twisted to fit where ever it is needed. The wires are also able to shift in length as a motor pulls on them to control each wires’ length individually. The software is able to take the desired shape and configure each wire to the right length in order to fit it.

The company believes that the invention will be useful in a wide array of fields including aerospace assembly, nuclear inspection, and even certain types of surgery. The technology is very easily manipulated which allows people to use the arms in a variety of ways. The company even claims that it can be used as a steerable hose pipe or vacuum cleaner, although the price will probably be too high for people to buy it for their households. Currently only the 24-inch arm pictured to the right has been shown, but the company says they are currently working on arms of different lengths.

The AI Landscape poster celebrates 50 years of Artificial Intelligence research and development

(Click on the image for a high resolution version.)

The AI Magazine's summer edition published by the Association for the Advancement of Artificial Intelligence (AAAI) is celebrating half a century of Artificial Intelligence research and development. Today, it is far more common to hear people talk about the failures of AI and all the promises it never fulfilled rather than hear about the successes of AI and how important its integration into our everyday life is going to be. The magazine hopes to promote AI by publishing the AI Landscape poster shown in the image at the beginning of this post.

It is true that we are far from having strong AI some time soon and that very little research today is undertaken with the goal of producing such a system. Most academic research is focused on a very narrow area within AI often producing small improvements over current methods of solving very specific problems. As a result, even though a general purpose AI is not available neither commercially or academically, some of the smaller problems solved in the last 50 years are quickly finding their way into everyday life. The AI Magazine's AI Landscape poster is an effort to showcase how AI research has advanced over the last 50 years and how it will become an important element in daily life in the not so distant future.

As you can see from the poster image at the beginning of this post, AI is making inroads in a number of human affairs such as education (robots for education, smart desks, tutoring systems, and efficient Internet search engines) and medicine (expert systems for diagnosis, smart wheelchairs, assistive systems for people with dementia, drug design and robotic surgery.) Researchers have also made huge leaps forward in term of designing robotic bodies similar to our own (consider for example the proliferation of humanoid robots in the last 5 years compared to the previous years when only Honda could afford to develop such a complex machine.) Some commercial robots have also emerged and have quickly become popular (an example is, of course, the very popular vacuum cleaning robot Roomba from iRobot and several museum tour guides that are making appearances in museum around Europe and North America since the late 90s.) Let us also not forget the scientific research taking place outside the confines of our pale blue dot allowed to us because of the creation of intelligent robots such as NASA's Mars Rovers and other interplanetary spacecraft.

Exoskeletons (see the Sarcos exoskeleton for example) are not mentioned but I think they should have been for no other reason but the immense potential of such machines to help differently enabled people.

Other areas of great advancement include hand written recognition, spam filtering (email would have been rendered useless as a medium of communication if it was not for AI,) fraud detection, vehicle navigation, and smart homes (not many of these around today but I bet in another 10 years more of us will be able to afford a large variety of automation systems in our homes.)

The goal of AAAI's poster is to spread awareness among the people about how AI systems have advanced over the last few decades and even prepare us for more applications that are sure to follow in the future. Help AAAI spread the word by discussing the poster. Together, we can eliminate misinformation about the true achievements of Artificial Intelligence and encourage more people to study but most importantly accept AI in their daily lives.

The AI Landscape poster is copyright AAAI; get the poster in full resolution here.

Improving video quality using high resolution digital photographs

Researchers at the University of Washington have devised a new algorithm for enhancing and editing video. Given a low resolution video of a scene, the algorithm takes advantage of high resolution digital photographs of the same scene in order to improve the quality of the original video. Some of the enhancements include improving the resolution of the video, dealing with lighting artifacts such as specular reflections, improving the video camera's dynamic range, and removing camera shake. Also, the algorithm allows users to seamlessly remove unwanted objects from the video and even edit the texture and color of select objects using a small number of actions, i.e., by only editing a few frames from the original video.

Given a video and a set of high resolution digital photographs, the team uses structure-from-motion and multiple-view-geometry methods to estimate a 3D representation of the scene. Using texture information, the team can estimate the location of the still photographs in the video and then blend them in the video correctly using the structure information and an image-based rendering algorithm. Currently, the method only works with video of static scenes.

The video below, explains the method and shows multiple examples of it used to enhance and edit several poor quality videos. I especially like the removal of objects such as the stop sign in front of the flower shop and the editing of individual objects such as the frames of the photographs in the last example.

In case you are wondering, this method is not real-time. In fact, it requires roughly 5 minutes of processing time per video frame of 853 x 480 resolution. Fortunately, this method is mostly automatic requiring very little interaction from a user. The researchers also suggest that the processing time can be significantly reduced if certain parts of the algorithm are optimized more carefully. There is more information about this video editing method at the project's research web page which includes a link to the published scientific paper.

New Mars rover revealed

UK based space company, Astrium has unveiled two new prototypes for future Mars rovers. The two creations are regarded as the most robust and maneuverable rovers built to date and are scheduled to be complete and sent to Mars by 2015.

Chris Draper, Astrium's ExoMars rover vehicle industrial manager, has said that the six-wheeled rovers can literally go where no rovers have gone before due to how easily they can adapt to different situations. Draper was very enthusiastic and confident about the release of the robots, claiming that they wil even be able to outperform the applauded American MER rovers, which outlasted theri expected lifetimes.

Astrium's rovers have six wheel steering, meaning that they can turn in more directions faster. They are also equipped with wheel walking capabilities, meaning that on uneven surfaces they can essentially anchor down with five of the wheels while slowly moving forward with the sixth. From the prototypes it looks like Astrium may be able to pull this off because the traveling and navigation software is quite impressive. In fact, the navigation system is powerful enough that the rovers can control themselves and explore without much human intervention.

The rovers will also have more scientific tools than any other rover currently built. This will mean more information can be gathered and since the rovers are expected to last a long time that could lead to big revelations about the Red Planet.

Although the final product is still several years down the road, the company seems optimistic with its technology and what it can do. The only obstacle that may remain is with funding, but if their prototypes are really that reliable, it shouldn't be long before an investor helps the company out.

Robot powered by biological brain

Scientists at the University of Reading have created a robot that is controlled by a biological brain composed of cultured neurons. Several other institutions have tried to achieve the same feat, but until now it has not been done. The developers are extremely proud of their creation, but the work is far from over.

The technology is going to be used in various studies of the brain. Currently, a project has been created that will study the how the brain stores memories and specific pieces of data. The long-term goal of the project is to learn more about neurological diseases such as Alzheimer's disease and how they occur.

So how exactly does this invention help with those studies? Well, in order to understand that one must learn how the robot and its brain work together. In simple terms, the cultured neurons are placed on a Multi Electrode Array (MEA), which communicates with the robot via bluetooth, sending messages from the brain cells. The MEA consists of approximately 60 electrodes that pick up electrical signals created by the cells. Currently, the "brain" is able to tell the robot when and where to move. When the robot nears an object the electrodes are stimulated and the brain knows that it is near something. The whole process is quite amusing.

The next part of the project (studying memories and data in the brain) will also use the robot as the scientists see how it remembers and reacts to an environment it has previously been in. By seeing the robots reactions and analyzing the brain cells, scientists will be able to get more information on exactly how the brain deals with data. After more on that is learned, the project will continue on to further studies of the brain and all along the robot will likely learn and grow just like us humans. Kind of cool isn't it?

PekeeII, Zig Zag Zog, Target Bot, and Sumo Bot: Robots by Wany Robotics

French company Wany Robotics has been in the robotics market for a decade. They develop mass market robots and related robotics technologies available for licensing and integration into third party systems such as intelligent vacuum cleaners. Wany sells a good range of mobile robots which cater to many people including the research and hobbyist communities as well as the general public and especially young children.

The most advanced robot Wany Robotics sells is the PekeeII which was released last May (an upgrade of their previous PekeeI offering.) The company sells 3 versions of the robot. The lower-end PekeeII Essential Package is targeted to the hobbyist market, although, the 3500 Euro price tag probably still puts it out of reach of the average person. The more affordable iCreate from iRobot is probably the dominant player in this market segment. Wany's PekeeII Advanced and Ultimate Packages add an on-board PC and more expensive sensor packs such as cameras and lasers to the Essential Package robot. The more advanced robots are more suitable for academic labs and industrial R&D centers which explains why prices go up to 5500 and 12300 Euro for the Advanced and Ultimate Package respectively.

In terms of software, PekeeII supports a large range of robot control architectures such as Microsoft's Robotics Studio, the open source Player/Stage platform, MATLAB, and URBI which I have never heard of before. Interestingly, purchasing the Karto Intelligent Mapping Software adds another 3200 Euro to the robot's price! In case you didn't know, there are several mapping packages available for free online today; for example, the Bayes++ library includes a free implementation of the FastSLAM algorithm.

Wany Robotics also makes a few toy-robots for children. Of those, I find the Target and Sumo Bots the most interesting. The first one can navigate avoiding obstacles and has little guns for shooting things. How exciting is that? The Sumo Bots are claimed on the company's front page to have real Artificial Intelligence but I would bet that this is probably a massive exaggeration of the actual AI available on these toys. Lastly, Wany makes the Zig Zag Zog robot which looks like the cutest of the bunch and probably the one targeted to the youngest of its audience.

Rubber that conducts electricity

Researchers at Tokyo University claim that they have developed a special type of "skin" that will allow robots to essentially feel. The creation is a type of rubber that can conduct electricity well, allowing it to feel heat and pressure. Up until now there has been no material like this because metals aren't stretchable and rubber can't conduct electricity well. This invention is bound to have a big impact on future technology if it is proven to be reliable.

The developers boast that the new material can conduct electricity up to 570 times better than commercial rubbers with carbon particles. The flexibility is just the same though, which could mean big things for the electronics world. New wires can be produced with the material that will be able to bend into almost any shape without breaking. According to early tests, the rubber is able to stretch up to 1.7 times its original size.

Another possible use being discussed is in robot development. The material would allow for robots to be much more responsive and life-like because their entire surface could have sensors since the electrical power source is everywhere. The day when robots match humans is even closer, as this material could be the "skin" of the robot human since it can be maneuvered to do virtually everything our skin does when it reacts with our surroundings. It seems quite surreal and perhaps scary, but we are closer to an "I, Robot" type world than some of us think.

Swarm robots may be the future

The Artificial Life XI conference was hosted by the University of Southampton this week where there were many swarm robots, some old and some new. Some scientists claim that swarm robots are much more adaptable when it comes to complex procedures and that means that they could be more prevalent in the future than today’s popular one-bodied robots.

Swarm robots are a group of small robots that each perform a small task, almost like an assembly line. Their combined efforts usually add up to a big task being completed. Since it is a lot easier to program several robots with simple tasks than it is to program one robot with a complex task, the idea of swarm robots certainly seems plausible. On top of that, the smaller and simpler swarm robots often cost a lot less.

A group of undergraduate students at Southampton actually created a group of swarm robots that only cost 24 pounds each to create. An all-in-one robot that was able to complete similar tasks would cost several times more. Also, it is much easier to diagnose problems in swarm robots than in big ones because the different parts of the process are right in front of the user’s eyes. Of course the replacement cost is also cheaper.

The hopes for swarm robots are certainly high among this group right now. There is talk that these assembly line robots will be more effective at exploring space and prototypes for such a robot are already in development by private groups. Some are also suggesting a co-existence between big robots and swarm robots.

Autosub6000 to start research work in the Caribbean

According to the BBC, UK's newest autonomous underwater vehicle known as Autosub6000 is on its way to explore the deepest undersea volcanoes in the Caribbean. The robotic submarine can dive as far as 6000 meters and it will be complemented by the remote controlled submarine Isis. The robots will explore the volcanoes all along the Cayman Trough with Autosub6000 going first to locate the volcanic vents at the bottom of the ocean and Isis going second to collect samples around the vents. The two submarines will help scientists learn more about life in the deep ocean; the team on board the vessel James Cook which will be heading to the Cayman Trough expects to find many new species during the expeditions.

Robots are helping us explore so much of nature we could never hope to explore ourselves in the near future. Several robots are currently studying the surface of Mars looking for water and traces of past or current life while these two submarines will help us understand life on our planet. The Lunar X-Prize is determined to prove that robotic missions to the Moon can be both cheap and an effective method for scientific research (not that the latter has not been proved already by NASA's planetary rovers but those were definitely not cheap.) Scientific applications for robotics are by far the most valuable and worthwhile of all robotics endeavors undertaken today.

Read more: Sub to make deep Caribbean dive.

Robot helicopter learns autorotation emergency procedure for landing

Andrew Ng's autonomous helicopter group is at it again. After using reinforcement learning to teach their robotic helicopter how to fly inverted, they have now also managed to teach it the autorotation emergency procedure for safely landing after an engine failure. This procedure is used by real-life pilots to safely land their helicopter using the craft's potential energy to maintain sufficient rotor speed for landing.

Whereas during powered flight, the rotor drag is overcome by the engine power, during autorotation the rotor drag is overcome by the airflow through the blades. Effectively the potential energy of the helicopter (corresponding to its altitude) is transferred to rotor speed. This rotor speed allows the pilot to control the helicopter throughout its descent, and then slow down the helicopter before touching down.

Last month, the Stanford team presented a paper at the International Symposium on Experimental Robotics (ISER) describing their method. The paper also won the Ben Wegbreit IFRR Student Fellowship Award. Needles to say, autonomous control of a helicopter is far more difficult than the control of fixed-wing aircraft or ground vehicles. Especially in the case of landing with complete engine failure, the helicopter pilot has only one chance of successfully executing the autorotation maneuver and avoiding disaster. During the last few years, the Stanford team has made huge contributions towards solving the problem of helicopter autonomous control.

There are several videos of the autonomous helicopter in action at the team's official web site here.

Nuvo source code for sale

ZMP Inc., a Japanese robot developer, is selling its source code for the biped robot labeled nuvo to the first 50 buyers. The organization hopes that the feedback they receive from its buyers will help them evolve the technology. The bundle, which comes with a nuvo robot, the source code, an electrical circuit diagram, training, and an ID for participating, will cost a little under $2,500. Since there are limited spots, it probably won’t be long before all of the packages are snatched up, but with the high price tag ZMP is almost guaranteed to get people committed to helping the cause.

The company believes that the technology has great potential and can be used for many real world applications. Currently, ZMP is developing versions of the robot for both the entertainment and education sectors. The “e-nuvo” is one of the industrial uses of the technology and is actually quite popular among automobile manufacturers.

In the announcement, the CEO of ZMP stated that they were reluctant to release the source code, but found it necessary in order to speed up the growth of the technology. Before that happens though, more research is needed and that is exactly why this move has been made. The first users of the program will be Professor Yoshiki Matsuo of the Computer Science Department at Tokyo University. He will be using the technology for education and research purposes. ZMP hopes to begin the diversification of the technology as it gathers more data and it feels confident that they have a solid foundation and a stable program that will allow its users to quickly send them suggestions and the data they require.

Remote-controlled pill bots

A group of scientists at Carnegie Mellon University has created a pill bot that may have a huge impact on the medical world. To recap, a pill bot is a tiny machine often attached with a camera that humans can swallow. Doctors use them to see what is going on inside the body of a patient. CMU’s new pill bot is an improvement on most of the current ones in use.

It comes with a pair of tiny, adhesive hands, which will allow it to stick to internal surfaces. The doctor will be able to control the robot’s arm, meaning that he/she could focus more on the part of the body that seems troublesome instead of leaving it up to chance. On top of that, the creators of the gut bot claim that it will also be able to deliver drugs and other treatment applications.

If everything works out as planned, this invention will certainly improve medical arsenals around the world. It will be especially useful for biopsies and combating internal bleeding with tiny built-in lasers. It almost seems unreal, but looking at the amount of time the group spent researching and developing the product, it seems plausible.

According to the developers, the toughest part of creating the pill bot was finding an adhesive that would stick to various tissues without ripping or damaging them. Adhesives that stick once alredy exist, but finding one that would stick and then “unstick” was much harder. With a lot of work the team was finally able to find a solution and have created a preliminary product. There is still a lot more testing to do and the technology with certainly be built upon, but the foundation is laid as the medical world eagerly awaits for yet another tool to help save lives.

Robots improve surgery

Robotic surgery is a growing trend around the world as it is proving to be more efficient and effective than manual surgery in many aspects. A robot is able to create much more precise cuts than a human can with a scalpel, which can often mean less bleeding and less discomfort for the patient. Machines can ensure that the cuts are no bigger than needed for the surgery to take place and they can also easily avoid nerves and blood vessels. Although this may not seem like a big deal, it really is.

The human hand isn’t very accurate compared to the machines and accidents are a lot more likely. On top of that, the incisions made by the robots can be as small as dime-sized, while a similar surgery by hand would probably end with a 4-inch cut. This would lead to a drastic difference in the amount of time the patient has to stay at the hospital for recovery. With the small robotic incisions, many patients are able to leave within 24 hours of the surgery, often with slim to no pain.

Surgeons across the globe are starting to grow fond of robots as they begin to use them more and more. Currently, robotic surgery is available for hysterectomies, certain heart procedures, and the removal of cancerous kidneys and prostates.

Although the robots are very useful, surgeons are far from having to worry about being replaced by machines. The surgery is still controlled by humans, and the robots are really just another tool in a surgeon’s arsenal. They are essentially just scalpels, but instead of being controlled directly by a human, they receive commands from a computer which the surgeon uses.

Bristled robots will help clean pipes

Scientists from the Dalian University of Technology have created a robot with bristles like a bottle brush. The creation is expected to become very useful over the next few years as it is able to go inside gas and sewage pipelines, which otherwise are unreachable at the moment. Currently, these pipes must be taken apart if they need to be inspected, but the bristled robot will eliminate the need to disassemble and reassemble the pipes, as it will be able to provide a wide array of data concerning the pipe’s condition.

For cleaning purposes, a robot called a “pig” is currently used. The pig flows along with the gas or oil inside the pipeline and it attracts and picks up any magnetic materials. The problem with this approach is that the [pig] robot won’t fit inside all type of pipes, particularly twisted ones. It also has a very hard time maneuvering through damaged structures. The bristled robot is much more compact and will be able to get inside obscure shapes, hence making it a lot easier to maintain the pipelines. It will also be more effective in picking up dirt and outside substances.

The bristled robots are sent into the pipes attached to a tether, which sends back data. There are different sizes of the robot to fit inside different pipelines, and so far in trials the success rate has been very promising (for gas, oil, and nuclear plant pipes.) With a huge demand and good test outcomes we could see the bristled robot become very popular among chemical and utility companies over the next few years.

Phoenix confirms existence of water on Mars

As far as completing the water finding part of its mission, NASA's Phoenix Mars Lander successfully verified the existence of water on Mars. According to a NASA news article Thursday, one of Phoenix's instruments detected traces of water vapor released after heating Mars soil that the spacecraft had scooped up from the ground using its robotic arm.

Given how successful in completing its mission so far Phoenix has been and also given that the spacecraft has remained healthy over the last few weeks operating on the surface of the Red Planet, NASA has also announced an extension of its mission for another five weeks over the duration of the original mission that was to be 3 months. This means that NASA will operate the vehicle until the end of September. I am willing to bet that unless the Phoenix Mars Lander becomes inoperable due to mechanical failures, NASA will further extend its mission even past September. The twin rovers Spirit and Opportunity are still exploring the Martian surface long after the datelines of their original missions have passed.

NASA's robot astronauts are by far the coolest robots in service today.