1.

INTRODUCTION

RAL (Rutherford Appleton Laboratory) provided the high-resolution camera for the UK TopSat satellite mission, launched in October 2005. The mission commenced full operations in January 2006 and the satellite is now routinely producing imagery. This paper presents some of the first results from the camera and describes the work that is presently being conducted in analysing these images.

2.

MISSION PARAMETERS

Orbit : 686km altitude, sun-synchronous polar orbit

Resolution : 2.85m panchromatic & 5.7m colour (at 600km altitude, 2.5m pan & 5.0m colour)

Swath : +/- 15km

Mass : 140kg including 29kg camera

Launched : 27th October 2005 on Kosmos-3M

Lifetime > 2 years

In order to increase the signal to noise in the detector, the satellite can perform a pitch compensation manoeuvre (PCM) to slow the ground velocity of the field of view of the detector through the image. This is shown pictorially in fig. 3 & fig. 4.

Fig. 1.

The TopSat spacecraft

Fig. 2.

The TopSat camera

Fig. 3.

PCM = 1, Image acquisition = 2 seconds

Fig. 4.

PCM = 2, Image acquisition = 4 seconds

The camera operates in a push-broom mode, with the camera focal plane assembly (FPA) having two linear CCDs scanned along the surface of the Earth by the motion of the satellite. One detector generates panchromatic data and the other three-band multi-spectral data.

3.

GROUND SAMPLING DISTANCE (GSD) ANALYSIS

An investigation of the achieved GSD has been performed using the image of Dartford, UK (fig. 6) taken on the 7^th December 2005 with a PCM factor of 4.

Fig. 5-

Pixel-to-pixel variation for uniform scene.

Fig. 6-

Dartford, UK, 7^th Dec ’05, 1030am, PCM=4.

Features with large separations were selected from the image, then the distance between points measured using an ordnance survey landranger 1:50,000 scale map (10.0m accuracy). Corresponding points were then located in the image data and pixel separation calculated. The effective GSD was calculated as the ratio of distance between points (km) to the distance between points (pixels). Only features completely along track or across track were selected. By selecting large distances between points, accuracy of GSD measurement is < 0.5%.

The across track results summarised in Table 1 match well with the expected GSD values (2.85m panchromatic and 5.7m colour). The along track GSD corresponds to a smearing effect of ~6%. This could be the result of small inaccuracies in the PCM.

Table 1–

Summary of GSD Analysis

4.

CCD OFFSET AND GAIN SETTINGS

The panchromatic and colour CCDs on TopSat have adjustable offset and gain settings that can be changed using a control file called a wavetable. These wavetables are stored in the Data Handling Unit and are uploaded prior to an image acquisition. The default offset and gain settings were establish by ground testing, but these needed to be confirmed once on orbit. So to this end some images were taken of uniform dark and bright scenes during the commissioning phase of TopSat. The figures in tables 2 & 3 are in engineering units or counts, with full scale equal to 1023 (i.e. 10 bit digitisation). The offset and gain for the images were 43mV and 1.0 respectively.

Table 2–

Dark Scene Image Statistics

Table 3–

Typical Urban Scene Image Statistics

SZA = Solar Zenith Angle

DR = Dynamic Range

s.d. = standard deviation

For the dark scene, the output from the pixels should be as close to zero as possible so that the full dynamic range of the CCD can be utilised. However, it can be seen from Table 2 that the pan, red and green channels are significantly above zero suggesting the offset for these channels could be reduced. The blue channel offset is about right so does not need to be adjusted by much. Unfortunately, due to the design of the CCD electronics, it is not possible to adjust the offset for the red, green and blue channels independently, so only a small decrease in the offset can be implemented before the blue channel would start to produce negative values. However, this small change in the RGB channel offsets might be sufficient to overcome some of the problems with saturation of the red channel (table 3) and so may be worth implementing. A reduction in the pan channel offset can be implemented quite easily by uploading a new wavetable. The dark scene image also gives some information about the level of noise from non-signal dependant sources, such as dark current and read-out noise. For the pan channel the noise is 1 LSB rms and the colour 0.5 LSB rms.

The data in table 3 shows that for imaging at relatively high latitudes (Dartford is at 51.5 deg N) with high solar zenith angles, a gain of 1 and a PCM of 4 utilises the full dynamic range of the sensor.

5.

FLAT FIELDING

To provide information about the pixel to pixel variation under constant illumination conditions, an image was taken over the Antarctic ice with a SZA of 60° and PCM of 1. Figure 5 shows how the response varies across the pan channel array. Data from this flat-field image will be used to develop de-striping algorithms for correcting images after they have been received on the ground.

6.

FUTURE WORK

Most of the images shown in this paper have been taken with the camera operating below its nominal operating temperature, in order to conserve power during the commissioning phase. It is known from ground based testing that the camera performance will improve when the heaters are switched on, so some of the scenes will be re-imaged and a comparison of image quality made. This work should help establish an optimum operating temperature for the camera before proceeding to make further performance measurements.

An evaluation of the camera’s in-orbit modulation transfer function (MTF) is also planned. The use of well characterised ground targets, such as black and white square pattern targets, is being considered for this purpose.

Other activities include assessing the pointing stability during high rate pitching manoeuvres and oversampling scenes to improve the along track GSD.

7.

EXAMPLES OF EARLY IMAGES

Fig. 7-

Oklahoma, USA, 22^nd Dec ’05

Fig. 8-

Northern Australia, 8^th Jan ‘06

Fig. 9-

Saskatoon, Canada, 15^th Dec ‘05

Fig. 10-

Washington, USA, 31^st Mar ‘06

Fig. 11-

Hakodate, Japan, 11^th Jan ‘06