Transcript
1 2
Uncooled Thermal Imaging
What is the future?
Shutterless Non uniformity correction Non-‐uniformity in uncooled thermal imaging systems is an undesirable characteristic arising from small differences in the responsivity of individual detectors. Non-‐uniformity typically manifests itself as FPN (fixed pattern noise) in the raw data image and becomes a significant problem since the sensor array outputs are also affected by the spatio-‐temporal temperature fluctuations of the FPA.
The most commonly used technique for non-‐uniformity correction (NUC) consists of estimating, through a calibration procedure using a shutter, an array of individual (gain and offset) correction factors applied to the outputs of each corresponding pixels in the image.
Size, weight, power and manufacturing cost of the infrared imaging system are increased because of the shutter. Every time the shutter is activated, the real-‐image image capture process is interrupted for a few seconds. So, the camera is blind for a few seconds
Pre-‐loaded calibration tables to create offset matrices combined with scene understanding algorithms to estimate non-‐uniformity based on scene radiance information.
Non-‐uniformity is a serious practical limitation to both civilian and military applications -‐ as it severely degrades image quality.
Tonbo Technology
Low power: Tonbo cameras consume 0.6W power for QVGA and 1W power for VGA resolutions). Comparable cameras (using shutter for NUC) consume twice the amount of power. For hand held systems, this provides double the amount of operating time on batteries Always on: Tonbo cameras are never blind and are always on because of shutterless operation No moving parts: Less chance of failure on the field
1 2
Application specific video processing on the camera Thermal imaging cameras are deployed on moving platforms and in areas with inclement weather. This causes severe disturbances in the captured video rendering them useless for analysis and also causes severe strain on the viewer
The video from traditional cameras is transmitted in its jittery form to a central server. At the server it is either post processed or processed inline to give stabilized video.
As the number of cameras that are deployed increase, it becomes very difficult to have an array of processing stations just to stabilize incoming video feeds. This adds to the cost, maintenance of these processing stations and also not scalable as the number of cameras increase.
Tonbo Technology
Tonbo thermal imaging cameras have built in electronic video stabilization. The output from the camera is jitter free and stabilized. If the camera is used in a moving scenario (either panning or on a moving vehicle), the onboard stabilizer preserves the intended motion while removing shake. The stabilization is a combination of image registration using a gyroscope and real-‐time multi-‐ resolution image warping all packaged inside a state estimation and filtering framework.
1 2
Megapixel thermal imaging
Uncooled thermal imagers rely on expensive material (ASi or VOx) for sensing incoming radiation. The cost of the material limits the extent to which detector resolution can be increased. Detector resolution is important because a higher resolution detector means a smaller size lens for the same field of view. Currently, the maximum (production ready) detector resolution is 1024 x 768 pixels. Custom detectors exist for higher resolutions. But they are very expensive.
Currently a 2-‐megapixel thermal imager resolution is achieved by using a custom detector with a native 2-‐megapixel resolution.
Imagers based on focal plane arrays (FPA) risk introducing in-‐band and out-‐ of-‐band spurious response, or aliasing, due to undersampling. This can make high-‐level discrimination tasks such as recognition and identification much more difficult.. Increase in native detector resolution leads to a significantly high cost.
Tonbo Technology
Tonbo technology takes multiple images of the same scene, according to different patterns while displacing each time the image over the detector plan by a distance equal to a fraction of the detector pitch. The under-‐sampled frames of the scene are then used to form a single high-‐resolution frame.
Elimination of deleterious artifacts from staring arrays Elimination of aliasing and spurious response Quantitative resolution improvement Qualitative Image resolution improvement Minimum Resolvable Temperature (MRT) Difference improvement
