We focus on nine societal areas.
TNO cooperates with companies, the public sector and other organisations, to apply our knowledge and expertise with and for others.
We develop knowledge not for its own sake, but for practical application.
TNO offers you the chance to do groundbreaking work and help customers and society with innovative, practical and smart solutions.
Existing patents describe a method for detecting moving objects from a moving camera, using a system that compares multiple frames. Subtraction between two images leaves objects as a remainder. This process can be somewhat messy, since the process of comparing two frames requires simultaneous interpretation of the background and of the object. The proposal involves using an artificially high-resolution image to organize the comparison between images. The inclusion of this synthetic image has the power to decrease the noise of the captured images.
The invention calls for image processing that, rather than compare one frame to the next, first creates a higher-resolution synthetic image based on multiple frames. Individual frames can then be compared to the larger image, which represents the background more accurately. A moving camera actually makes the creation of the synthetic image easier, because it creates more data about the background—it creates variation in the range of visual reality described by each pixel. Once the background is established, the system can more easily identify, track, and image a given moving object.
Multiple frames collected from an input source are synthesized into a higher-resolution background image using previously known techniques. Several options are available for this synthesis, such as the use of even or odd images. The associated computer program will use an algorithm that identifies a certain threshold for object detection; if pixels differ in this degree from the background composite, the program will register the existence of a moving object. The computer program can then also track the object across multiple frames and/or yield an extracted image of the object.