A sampling of images produced with the GMR sensor array is shown in Figures 4-7. Figure 4 shows an image of a #10 threaded rod 1.5 cm above the array, as pictured in the top-view drawing on the right. The two gray-scale images on the left are data obtained from the sensor array. The top image shows data from the GMR elements (labelled "corners") with a vertical axis of sensitivity. In the "corners" image, a value of 0 volts is displayed at the positions corresponding to the elements with a horizontal axis of sensitivity. The bottom image, labelled "others," shows data from the GMR elements with a horizontal axis of sensitivity. In the "others" image, the positions corresponding to the "corners" elements are displayed as 0 volts. Comparison with Figure 2 will clarify which pixels are meaningful in the two images. In a real UXO detection system, more sophisticated software would combine the two images in a contour or vector plot. Here darker grays indicate higher magnetic flux, while lighter grays indicate lower magnetic flux. The scale is on the right of each image. The numbers next to the grayscale are the sensor signal in volts, so that 1.4e-2 means 14 mV of signal.
In the top image of Figure 4, the magnetic poles on the ends of the rod are being picked up by the sensors at the upper right and lower left corners. In the bottom image, the sensors are responding to magnetic flux leaking from the sides of the rod. While the characteristics of the rod are not completely clear with this low spatial resolution, its general shape and size of the rod can readily be determined.
Figure 5 shows another image of the same rod, only this time flipped over so that it is pointing towards the opposite corners of the array. The movement of the magnetic poles to the upper left and lower right corners is obvious in both images. There are several reasons why this image is not a perfect mirror of Figure 4, namely different lateral placement of the rod on the array and different rotation of the rod about its own axis. The magnetic domains in the rod may not be azimuthally symmetric, with the result that the image may depend somewhat on which side of the rod is facing downward. The rod in Figure 5 is also oriented differently with respect to the earth's field than in Figure 4. In a real UXO system possible ambiguity created by different remanent states of objects can be addressed through application of a rotating alternating-current magnetic field created by two orthogonal sets of coils. A field large enough to force magnetic poles on each surface of a permeable object will make each surface visible to the GMR array, which is in essence a magnetic edge detector.
Figure 6 now shows an image of the same rod in a constant 6 Oe external applied field which was generated with air-core Helmholtz coils. Before acquiring the image, the background of the array was characterized in the presence of the 6 Oe field. The most striking part about this image is that it looks much like Figure 4, showing that the GMR sensor array is able to image ferrous objects even in the presence of a substantial background magnetic field (about fifteen times the earth's field). The reason for the similarity of the two images is that the array detects spatial magnetic field variations, not the scalar magnetic field amplitude like a cesium-vapor magnetometer. Figure 4 supports the assertion that array-based detectors will usable with magnetic soils as long as the ground clutter is reasonably homogeneous on the length scale of the objects to be detected.
Another point about Figure 6 is that the image is a bit clearer than in Figure 4, where no external field is applied. The improved image quality occurs because much of the flux from the applied field passes through the magnetically soft rod. A more complete outline of the rod could be made by acquiring another image with an applied field in the orthogonal in-plane direction. In fact, a real UXO system would likely incorporate 3 sensor arrays, each with a different orthogonal axis of sensitivity. Data would be read out from each array while a coil applying a magnetic field along that direction is energized. A fully realized system would incorporate a rotating magnetic field and synchronous acquisition from the 3 orthogonal arrays. An additional group of 3 sensors could be used with a portable GMR detector to eliminate noise due to motion of the detector in the earth's magnetic field. (Such a noise-elimination scheme has recently been described for a fluxgate vector magnetometer system by Allen et al.[Allen, 1996])
Figure 7 illustrates the ability of the array to image slightly more complex objects. Here two bolts have been placed 1.5 cm above the array. The "corners" image shows substantial flux from the threaded part of the bolt, while the "others" image shows a more difficult to interpret pattern of flux possibly arising from complex domain patterns in the bolt head. The overall "V" symmetry of the objects is apparent in both images. The image of the two bolts would be greatly improved by application of a rotating external field and by a higher resolution array, with more pixels on each object.
Another study was done to follow the evolution of images as a magnetized object (here a length of 3/4" rebar rod) is moved away from the array. When the rebar was close to the array, it blinded the detector, saturating most of the sensors. As the rebar was moved from 1.5 cm to 9 cm separation, the signal level was reduced from 280 mV to 210 mV, still well above the typical background level of 20-50 mV. The two images at different separation (not shown) demonstrate that we have an object with a vertical axis of symmetry. The image formed by the sensors with their axis of sensitivity parallel to the rebar axis has the same qualitative features independent of spacing. In contrast, the image with the orthogonal axis of sensitivity varies dramatically as a function of spacing. This variation is due to different spatial falloff of the various multipole components of the magnetization pattern. One must keep in mind that the magnetic field emanating from an object can vary in all three dimensions, and there is no particular reason in the absence of an external applied field for the symmetry of a 2D slice taken at one height to be exactly the same as a 2D slice taken at another height. On the other hand, the images sometimes appear rather simple, as in Figure 4. Intelligent synthesis of data and interpretation of images will be the major challenge in building a useful GMR-based UXO detector, although the intrinsic difficulty is not greater than in time-domain analysis of pulsed electromagnetic induction data, for example.