Background

Consideration of the physical environment is critical for wheelchair users. The most effective rehabilitation outcomes are realized when evaluations consider both functional and environmental factors1. While many tradesmen are able to perform the modifications, the availability of skilled professionals able to assess the home environment and identify what changes are necessary to meet the wheelchair user's needs is very limited. Providing services in rural areas is particularly difficult. A system that allows accurate remote assessments is an important tool to address this issue. The use of virtual reality technology and telerehabilitation concepts to assess the home built environments of persons with severe mobility impairments was recently proposed by University of Pittsburgh researchers2. The VRTS discussed here was developed in response to the proposed concept.

The VRTS uses a commercial software package, Photomodeler Pro 4.0 (Eos Systems Inc.) to transform a series of digital images of a space into a corresponding 3D model. Previous work showed that Photomodeler® has a high precision value of 2800:13. Usability is a primary concern for the VRTS4. In particular, the techniques and logistics involved in acquiring the images are critically important. The process of generating the 3D model from the images is somewhat labor intensive in that it takes a trained individual about 2 hours to generate a model of a typical interior room with 4 walls. In order to limit the number of visits to the remote site, it is desirable to develop an image acquisition protocol that can be performed by an untrained individual without direct supervision. It is therefore impractical to require the use of expensive and/or sophisticated camera equipment. Such equipment would either be too complicated to use effectively without training or too valuable to risk loss or damage by sending it to the site for use without supervision. To overcome this problem, we proposed that inexpensive disposable cameras be used on site by untrained individuals--either the consumer themselves or a caregiver. To study this alternative, we initiated the current study that compares the modeling accuracy using digitized images from the disposable cameras to the accuracy of high and low resolution images from digital cameras.

Research Question

Are either disposable film cameras and/or low-resolution (1.2 mega-pixel) digital cameras comparable to high-end, high-resolution (3.3 mega-pixel) digital cameras for creating accurate 3D models of interior residential spaces?

Methods

We compared four camera types for use in the VRTS, a disposable camera (Giant Eagle, 1.5 mega pixel photo CD scan), an inexpensive consumer level digital camera (Canon A10, 1.2 mega pixel), a high resolution digital camera (Canon G1, 3.3 mega pixel), and the high resolution digital camera (Canon G1, 3.3 mega pixel) with a wide angle lens Canon Wide Converter WC-DC58, 0.8 x wide) to apply the image acquisition and 3D modeling procedures to an actual practice with four cameras, the bathroom of a wheelchair user's house was used as a target environment. The test was executed as follows:

Results

Table 1 shows the comparison of measurements of 10 target dimensions taken from 4 different 3D models of the target bathroom. Deviations of each object's measurement between the tape measure measurement and in each 3D model were calculated. The model generated with the images from the disposable camera showed the lowest precision, 39:1. The models generated from the A10 and G1 cameras images showed produced precision values (59:1 and 63:1 respectively). The model generated from images from the G1 camera with the wide-angle lens showed the most accurate measurements of highest precision, 200:1. The models generated from the disposable, G1, and A10 cameras used 7 photos. The model generated from the G1 with the wide-angle lens used 6 photos. Figure 1 shows a 3D model generated from disposable camera.

Conclusions

Figure 1

  We have determined that disposable film cameras can be used in the VRTS. The model generated using the disposable camera was only slightly less accurate than the models generated with the high-resolution, digital camera. The accuracy of the model generated with the low-resolution digital camera also compared well with the high-resolution camera model. The best accuracy was obtained with the high-resolution digital camera with the wide-angle lens. This may have been due to the larger field of view resulting in better images from which to generate the model. The person performing the modeling noted that the images from the high resolution G1 camera were easier to use in the modeling procedure and therefore required less time to process than the images from the disposable or low-resolution cameras. Although the disposable camera produced less accurate models than the other camera configurations, it is still likely sufficiently accurate for assessing wheelchair accessibility. Moreover, the disposable camera has the advantages of affordability and ease of use.

References

Acknowledgements

This work was funded by the Department of Veterans Affairs and the University of Pittsburgh. Opinions expressed are those of the authors and do not necessarily reflect those of the funding agencies.

Jongbae Kim
University of Pittsburgh,
Department of Rehabilitation Science and Technology,
5044 Forbes Tower,
Pittsburgh, PA 15260,
412-383-6581,
jbkim@pitt.edu