Computer vision

The surface-reduction-gauging experiment proves the ability of laser-spot assisted, 3D image analysis to characterize the geometry of the surface and provide the CSM target with high precision. It also discloses the extent of measurement precision of surface-reduction gauging and reveals the relationship between the density of laser spots cast on the surface, in spots per unit area, to the accuracy of the characterized geometry of surface. The experimental results also prove the following three premises of the surface extent application of laser spots using CSM-based 3D nominal World-frame-coordinate estimation: Though the nominal optical model and nominal robot kinematics model in the CSM system are globally imperfect, the premise is that as long as the laser spots are close enough to each other within the asymptotic-limit region or volume the relative error between them is close to zero-mean. Therefore, any error on the CSM side does not propagate into the measurement of surface reduction relative to an as-located original surface. The experiment repeatedly proves the premise for a range of surface positions and orientations. Another premise is that error in thickness assessment with laser-spot data is unbiased and random. This means that, provided they are matched, a laser-spot center detected in each camera corresponds to one single physical juncture on the object's surface. In other words, only error in thickness assessment with laser-spot data can be averaged out by applying a large amount of laser-spot data on the surfaces. The results of the above experiment repeatedly verify this premise. There is no bias between 2D position of the detected center of a circular cue and that of a laser spot in camera space if they occupy same physical location. (The "physical location" of the laser-spot center is not, strictly speaking, actually defined. What the tests indicate is that, on average, the software places each camera's camera-space center of a given laser spot in such a way as to represent the same surface juncture in all three cameras.) High-precision surface-change gauging extends the vision-guided-robot system based on CSM into a more potent surface-operating system, one that in practice only can be done by humans, and this in an imprecise, non-uniform, and often ineffective way. A number of surface-finishing tasks entail application of force or pressure combined with in-plane motion in order to achieve a scrubbing, polishing or sanding effect. Such tasks often make use of human dexterity and effort, which can result in repetitive-motion injury and incomplete or uneven treatment of the surface. But with high-precision surface reduction gauging and The general GCM model introduced in this chapter includes the best features from conventional camera models CCMs and models specialized for wide angle fisheye cameras in one unified model structure. Therefore there is no longer a need to use different models for these camera types. The GCM may also handle varying entrance pupil and decentring distortion as well as variations in focus and zoom. In an experimental investigation it was concluded that the accuracy using the new radial distortion compensation in GCM performed better compared to the CCM model tested when few camera parameters were used. The additional distortions in GCM can be added for increased accuracy depending on the camera used. The decentring distortion is usually more important to include in models of cheap cameras, while "entrance pupil" variation can be important even for high precision cameras and mostly if it has a large viewing angle, and is more important for variable zoom and focus applications. For future research we recommend to perform the accuracy investigation for different kinds of cameras, especially fisheye cameras. With a larger angle of view the advantage of the GCM should be bigger. These experiments should also be done with better leaning calibration images, so the compensation in Section 3.4 is not needed. Investigations on which kind of decentring distortion compensation is the best to use in practice should be done in such an investigation. It would be interesting to try the different "pre processing algorithms" e.g. with circular references, both for analysing how good parameter estimations can be achieved and also the accuracy of the "centre points" and calculation time. Another topic is to use natural corners in the environment as references. Also work on calculating analytical derivatives in the Jacobian for one or both of the transformation directions can be useful. To make calibrations with variable focus and zoom can also be of interest when the focus and zoom can be measured In this research, we developed two kinds of pilot assistance systems for T-52 and T-53 working in a disaster district. One is an interface that provides the information from several sensors efficiently for the pilot. Another is an interface that tracks and displays the distances of the points that are set by the pilot. This system can tell the pilot the distance by the change in the color that understands intuitively easily, and the processing speed is also fast enough. The pilot switches the two kinds of proposed interfaces according to the situation. The former is suitable for searching of a target object by the camera installed in the robot arm, and the latter is suitable for the operation that grasps a specific position of the target object by the arm. Then, the basic experiments for the construction of each system were conducted, and the effectiveness of them was confirmed. From now on, a further various tests and examinations will be necessary to use of these system and interface efficiently in the disaster scene. Moreover, more interface development research according to disaster scene to achieve the many kinds of work by the rescue robot will be necessary. We are continuing the research for the solution of these research topics now, furthermore have started the development for automatic work system of the robots