UltraCam Osprey Prime - A Sensor for Nadir and Oblique Imaging

Michael Gruber and Wolfgang Walcher of Microsoft’s UltraCam business unit describe the qualities of the latest Osprey aerial camera.

UltraCam Osprey Prime is the second generation oblique aerial sensor from Microsoft’s UltraCam business unit. It was first introduced at the ASPRS in Louisville, Kentucky in March 2014. The design goal – to combine a high quality metric nadir camera and large 45° oblique sensor heads – was modified slightly from the UltraCam Osprey, its predecessor. The upgraded nadir subsystem produces a panchromatic high-resolution image, a true-colour RGB image and a near-infrared image. The four oblique true-colour camera heads are now mounted at an off-nadir angle of 45° and deliver images from all four directions – forward, backward, left and right with respect to the flight path.

The most important design change is the focal length of the sensor system and the new arrangement of the oblique camera cones. The new nadir and oblique lenses show longer focal lengths namely 80 mm for nadir and 120 mm for the oblique camera heads. Thus aerial missions can be conducted from a higher flying level to provide the same ground sampling distance. From an altitude of 1000 metres the camera acquires a nadir ground pixel of 7.5 cm. The other important improvement comes from a newer generation CCD detector array which is used for the oblique camera heads and delivers 60 Mpx at stunning quality. Fast read-out of the detector arrays means that the camera can achieve a 1.8 sec interval between frames, increasing the productivity of the flight mission.

Camera design

The nadir part of the camera collects a high-resolution panchromatic image, true colour and near infrared images. The 45° oblique cameras are designed as single cones and assembled in their viewing direction so that optical prisms are not required.

Productivity increases in the air is one design goal. Using an overlap of 60% in flight direction and 40% between flight-lines the flight plan shows a base length of 225 metres and a line spacing of 525 metres which gives adequate overlap between the oblique footprints. At a flying altitude of 1000 metres and a nadir GSD of 7.5 cm the nadir looking image’s outermost extent covers an area of 2800 × 2800 metres.

The focal length of the four oblique camera heads is 120 mm and fits well to the specs of the nadir camera subsystem. At the 45° viewing angle the object distance is 1414 metres and the GSD varies between 7 cm and 10 cm in the image centre of the oblique images. The oblique cones are equipped with 60 Mpx CCD sensor arrays at a 6 µm pixel size and Bayer Pattern colour filters. The positions of the CCD sensor for the left and right looking heads are in portrait mode and slightly off centred. This gives more cross-track overlap but keeps the outermost footprint extent the same for all four oblique camera heads.

Camera calibration

Each individual cone is calibrated by means of a laboratory calibration in order to compute the parameters of the interior geometry of the cone as well as the lens distortions. This procedure is based on a set of photos taken of a three-dimensional target which is built up in the calibration laboratory. This target has a large number of well surveyed target points covering a field of view of 110°. In order to allow for the larger field of view of the new sensor, the target was expanded. Additional control points are available to cover the oblique cameras to the left and to the right. Tilting the camera by 90° also enables calibration of the forward and backward sensors.

The most specific characteristic of the calibration procedure is the simultaneous calibration of the nadir and the oblique camera cones. The quality and reliability of the calibration procedure for nadir camera heads is well proven. The nadir cone calibration provides the foundation for calibration of the oblique camera heads and therefore the exterior orientation parameters of those nadir cones are introduced into the setup for the oblique camera orientation parameter computation. This approach improves the redundancy of the adjustment as well as the quality of the eccentricity parameters.

The calibration procedure for the oblique cameras is based on a large number of images taken from 84 shot positions which are a combination of three camera stations and 28 different rotation angles. Thus a highly redundant set of image positions can be automatically measured and introduced into a least squares bundle adjustment procedure. The entire set of measured image positions consists of approximately 10,000 positions taken from the panchromatic nadir camera component and almost the same number of image positions from the four oblique camera heads. The quality of the calibration procedure is at the sub-pixel level.

The calibration procedure for the UltraCam Osprey Prime camera also includes radiometric calibration of the individual camera heads. The colour performance of the nadir colour cone and the four oblique cameras are derived and the respective colour correction matrices are computed. For complete calibration we also determine the behaviour of the individual shutters, offset and gain values of all CCD sensor arrays, as well as the lens vignetting at all supported aperture settings.

Image quality

The sensor technology of the new 60 Mpx CCD sensor arrays supports a dynamic range capable of delivering a digital signal at 12+ bit level. Thus each colour band of the resulting images can resolve more than 4000 distinct intensity levels.

In the spring of 2014 the first UltraCam Osprey Prime images were acquired for Microsoft-internal mapping projects. Results from aero-triangulation, surface modelling and ortho image production were analysed and three dimensional photo-textured models of urban environments were produced. The success of these missions shows the usability and high quality of UltraCam Osprey prime images. This shows the high potential of the UltraCam Osprey Prime, the well-balanced design and the benefit of the high-quality metric nadir sensor component.

This article was published in Geomatics World September/October 2014