Computer Vision News - January 2017

it and communicates it to the cloud, which in turn integrates input from several devices, some of which are not image-related like smoke detectors, door people counters and so on. These can use infrared beams to sense persons and (even without being image-related) they supply useful data. All relevant data is fused to generate a regular pattern , which is used to detect any deviation from it. This kind of investigation could not be done before the advent of IoT . Another example is given by telemedicine devices which are designed to monitor our health. Body conditions and ailments cannot always be described in a binary way (healthy/unhealthy): it might sometimes be necessary to take a picture of a wound, activate image analysis to obtain relevant information, send it to the cloud to get a response and even automatically connect to a remote doctor which will give the diagnosis. This example might be further expanded: when a drug is sent to the patient’s house, the pharma company producing it will be notified of the diminished inventory in the distribution channel, influencing its decision to produce another batch of that drug. This would work in a very similar way to the IoT application which tells what food inside a refrigerator has fallen below the desired supply threshold and at the same time informs the manufacturer that shops might soon need to place a new order of that food. Some decisions can even be taken at the cloud level, based on technologies like deep learning : properly instructed and properly trained, the system will have a sufficient knowledge of common treatments (possibly as suggested by doctors) and will identify them as a pattern: for a specific wound, the IoT application knows that doctors generally prescribe a specific treatment. When critical, it can even notify the closest hospital, sending an alert to the interested services (urgencies, ambulance, stretchers and the like) so that they can coordinate the arrival of the patient. If one wants to run IoT even before every device is equipped with this small “brain” (the CPU), the solution is image analysis . For example, every garbage bin in urban areas will integrate a device which will monitor its filling rate and eventually notify it to the garbage truck, so that it can come to collect. Whenever that was impossible to do, a regular camera can be mounted on the bin to monitor the trash level in it and notify the truck. This example shows how image analysis is closing the gap between the state of the art IoT technology and the practical conditions in the real world, with its constraints and limitations. As experts in basic image analysis, RSIP Vision’s engineers perfectly know how to get data from cameras, create sophisticated algorithms with small footprint and solve problems detected by IoT cameras. Together with that, our experience in deep learning analysis enables us to create efficient decision systems and tailor the perfect solution for the problem at hand. Project Computer Vision News Project 13 If one wants to run IoT even before devices are equipped with this small “brain”, the solution is image analysis