Introduction

Remote sensing is becoming an increasingly popular tool in precision farming applications. Digital imaging systems have distinct advantages over standard color infrared photography. Digital systems have tended to be prohibitively expensive. In this presentation, we evaluate the performance of an inexpensive digital imaging system in comparison to a research-grade remote sensing system.

The Systems

The two imaging systems compared in this study are the Dycam Agricultural Digital Camera (ADC, Fig. 1) and the Shafter Airborne Multispectral Remote Sensing System (SAMRSS, Fig. 2). The ADC is a commercially available digital camera that includes a modification to obtain images in the red and near-infrared (NIR) spectral wavebands. The cost of the Dycam ADC used in this study is less than $3000. SAMRSS is a custom-built multispectral remote sensing system based on three high-performance Dalsa digital cameras (insert in Fig. 2). The lens of each camera can be fit with a separate bandpass filter. This comparison will involve only the SAMRSS cameras fit with red and NIR filters. The cost to construct SAMRSS was approximately $70,000.

  Fig. 1.    Fig. 2.

The Comparison

Both systems were flown on two dates (July 7 and Sept 11) during 2000 aboard a Cessna light aircraft. Imagery was obtained nearly simultaneously by both systems of the same target. The aircraft was flown at 1524 m (5000 ft) AGL. Imagery from the two systems was compared in terms of its spectral, spatial, and contrast resolution.

Spectral Resolution

For the ADC, the red (635-667 nm) and NIR (835-870 nm) wavebands are similar to corresponding Landsat TM bands. The ADC wavebands are generally similar to the red (655-665 nm) and NIR (830-870 nm) bands used in SAMRSS.

Spatial Resolution

An ADC image is composed of 496 X 365 pixels, while a SAMRSS image is composed of 1000 X 1000 pixels. The relative sizes of an ADC and a SAMRSS image, on a unit pixel basis, are shown in Fig. 3. Figure 4 shows portions of the ADC and SAMRSS images in Fig. 3 blown up to a similar size (500% zoom factor for the ADC, 200% zoom factor for SAMRSS) using nearest-neighbor resampling (the portion from the ADC image is on the left, while the portion from SAMRSS is on the right). To judge the scale, the white square in the upper left corner of each image is 32 ft on a side. Thus, for images obtained at a given altitude, one would expect to see more detail in the SAMRSS image. However, overall image spatial resolution for a given system is determined by altitude and lens focal length. For a square target (like an agricultural field) that just fills the image, the maximum number of pixels that the width of the target would be resolved into would be 365 for the ADC and 1000 for SAMRSS. If the target were a field comprising a quarter-section, then the average spatial resolution of the ADC image would be 2.2 m (7.2 ft), while the average spatial resolution of the SAMRSS image would be 0.8 m (2.6 ft).

Fig. 3.

  Fig. 4.

Contrast Resolution

The individual red and NIR images produced by the ADC are 8-bit, while images produced by SAMRSS are 12-bit. Thus, the variation in scene brightness captured by the ADC can be resolved into a maximum of 256 levels. The variation in scene brightness captured by SAMRSS can be resolved into a maximum of 4096 levels. Thus, if the systems are optimally adjusted with respect to lens aperture, exposure time, and electronic gain, SAMRSS should be much more effective than the ADC is detecting subtle differences in scene brightness that might be associated with features of agricultural relevance. Figure 5 shows cumulative probability distribution functions (CPDFs) of brightness levels ("digital counts" or DCs) from identical portions of images obtained with the ADC and SAMRSS. Image brightness in each spectral waveband is spread over a much narrower range of DCs for the ADC as compared to SAMRSS.

Fig. 5.

Overall Image Quality

Red and NIR images obtained by the ADC on the two observation dates are shown in Figs. 6 and 7. Corresponding images obtained by SAMRSS are shown in Figs. 8 and 9. The SAMRSS images have been compressed to 8 bits and their contrast histograms have been matched to those of the corresponding images obtained with the ADC. All images have been scaled to a similar size. In general, the same major surface features appear in the images from both systems.

  Fig. 6.

  Fig. 7.

  Fig. 8.

  Fig. 9.

Conclusions

As might be expected, the performance of the Dycam ADC is not equivalent to that of the research-grade imaging system SAMRSS. However, it is our subjective opinion that the performance of the ADC, in terms of spectral, spatial, and contrast resolution, is acceptable for many agricultural remote sensing applications. Due to its relatively low cost and ease of operation, the ADC would be an affordable replacement for standard color infrared photography in many agricultural imaging applications.

Disclaimer

The results of this study are intended to provide useful information to those interested in this subject. The results of this study are in no way intended to represent an endorsement of any product by Texas Tech University or the Agricultural Research Service of USDA.

Return to MainPage