Transcript
Vumii Application Note CAMERA PERFORMANCE CRITERIA The Origin of Johnson’s Criteria At the first ever Night VIsion Image Intensifier Symposium in 1958, John B. Johnson, a night vision scientist at the U.S. Army’s Night Vision and Electronic Sensors Directorate (NVESD), presented a paper named the “Analysis of Image Forming Systems”. Johnson’s paper defined the criteria and methodology for predicting an observer’s ability find and assess targets using image intensifying equipment, under various conditions. Although newer methodologies exist, such as NVESD’s Night Vision Image Performance Model (NV-IPM), this criteria became the de facto standard in the defense industry for many years and is still used in the security industry today. Johnson’s Criteria Definitions Johnson’s model provided definitive criteria for calculating the maximum range at which “Detection, Orientation, Recognition, and Identification” could take place (with a 50% probability of success). To be concise, we have omitted “Orientation” from this application note as it is much less widely used and recognized . Detection Johnson defined “Detection” as the ability to subtend 1 TV line pair (+/- 0.25 line pairs) across the critical dimension of the subject (this translates to 2 pixels when using an LCD monitor). At the range that this occurs, regardless of target type, the observer could detect that a subject was in the field of view, 50% of the time. Today many security camera companies loosely follow Johnson Criteria and define their cameras “Detection” performance range as the ability to subtend either 1.5 or 2 pixels across the critical dimension of the target using various target sizes. Recognition Johnson defined “Recognition” as the ability to subtend 4 TV line pairs (+/0.8 line pairs) across the critical dimension of the subject (this translates to 8 +/- 1 pixels when using an LCD monitor). At the range that this occurs, regardless of target type, the observer determine the type of subject, a human or a car for example, 50% of the time. Today many security camera companies typically define their cameras “Recognition” performance range as the ability to subtend 6 pixels across the critical dimension of the target using various target sizes.
Identification Johnson defined “Identification” as the ability to subtend 6.4 TV line pairs (+/- 1.5 line pairs) across the critical dimension of the subject (this translates to 12 +/- 3 pixels when using an LCD monitor). At the range that this occurs, regardless of target type, the observer could detect the subject. Today many security camera companies loosely follow Johnson Criteria and define their cameras “Identification” performance range as the ability to subtend 12 pixels on target using various target sizes.
2014v1 DETECT RECOGNIZE IDENTIFY
Detection 2 Pixels Across Subject
Recognition 6 Pixels Across Subject
Identification 12 Pixels Across Subject
[email protected] | www.vumii.com
Johnson’s Criteria in the Security Industry Unfortunately, the application of Johnson’s criteria varies widely across the security industry and in most instances, documentation in the industry uses simplified or modified versions of the criteria. Typically, most companies use 12 pixels on target for identification, 6 for recognition, and 2 for detection (sometimes 1.5). However, the target size can vary greatly. Normally the defense industry “NATO” target is used for calculating the performance range for detecting vehicles, which is 2.3 x 2.3 meters but for a human target, various target sizes can be found. In any given document the target size for a human can range from 1.7 to 1.83 meters tall and from 0.3 to 0.75 meters wide. The Need for Better Performance Because end-users often place a high value on the written specifications of the camera, marketing departments are under pressure to use performance calculations that make their camera look better than their competitors. However, since these calculations typically do not take into account weather, environment, and camera sensitivity factors, two cameras that have the same pixel size and lens focal length, should always calculate to the same performance ranges. To achieve better performance calculations, company’s may use larger target sizes, smaller pixel counts (such as 1.5 pixels for detection, and/or use the height of the subject (and not the width).
Vumii Imaging In 2014, Vumii revamped its performance calculation methodology based on new criteria that is also verified through empirical testing. In all documentation, Vumii uses the standard 12 pixels on target for identification, 6 for recognition, and 2 for detection with target sizes of 1.8 x 0.6 meters for a human and 2.3m x 2.3m Focal Length x Target (L x W ) on a vehicle. However, Vumii believes that a square target more Performance Range = Pixels on Target x Pixel Pitch accurately portrays a real life scenario and so we use the square root of the product of the length multiplied by the width of the subject.
Example: Vumii Sii AT Thermal Camera Performance Range =
Detector pixel size: 25 micron Detector resolution: 640 x 480 Lens: 100mm
=
100mm x (1.8m) x (0.6m) 2 x 25micron
=
100mm x 1.04 2 x 0.025mm
=
104mm 0.05mm
= 2080m
2014v1 DETECT RECOGNIZE IDENTIFY
Focal Length x Target (L x W ) Pixels on Target x Pixel Pitch
[email protected] | www.vumii.com
