Close-range Photogrammetry Of Non-metric Digital Camera In Engineering Applications

Close-range photogrammetry has found many diverse applications in the fields of industry, biomechanics, chemistry, biology, archaeology, architecture, automotive, and aerospace, as well as accident reconstruction. Although close-range photogrammetry has not been as popular in bridge engineering and tunnel engineering as in other fields, the investigations that have been conducted demonstrate the potential of this technique. The availability of inexpensive, o?-the-shelf digital cameras and soft-copy, photogrammetry software systems has made close-range photogrammetry much more feasible and a?ordable for engineering applications. To increase awareness of the use of this powerful non-contact, non-destructive technique in engineering monitoring field, this paper presents a literature review on the basic development of close-range photogrammetry and brie?y describes previous work related to project deformation and geometry measurement; structural test monitoring, and historic documentation. The major aspects of photogrammetry project measurement are covered starting from the late 1970s and include a description of measurement types, cameras, targets, network control, and software. It is shown that early applications featured the use of metric cameras (specially designed for photogrammetry purposes), di?use targets (non-retro-re?ective), stereoscopic photogrammetry network layout, and analog analytical tools, which transformed over time to the use of non-metric cameras, retro-re?ective targets, highly convergent network layout, and digital computerized analytical tools.The thesis studies and makes the digital close-range photogrammetry based on the non-metric camera in consideration of the metric camera is high-cost and complicated. MATLAB neural network and has the powerful self-adaptive and self-learning capability and can be used to deal with the system which mathematical model is hard to describe. This paper mainly studies on the basis theory and method of realizing non-metric digital camera three-dimensional (two-dimensional) measurement based on MATLAB neural network, and applies the results to the deformation measurement of bridge and tunnel.Based on the direct 2-dimention linear transformation mechanism of close-range digital photography and mathematical principle of the linear neuron, the equivalent relationship between linear neural network and 2-dimention DLT of close-range digital photography is discussed. A neural network with 2 linear neurons, 6 inputs and 2 outputs is established to simulate the 2-dimention DLT. The network can be trained using a set of grid points in the control coordinate system with known world coordinates and pixel coordinates. The weights and biases of trained network contain camera interior and exterior parameters. A new digital photographic technique is put forward combined camera self-calibration based on neural network with non-linear pixel coordinates correction of lens distortion. Meanwhile, the new technology is used in crack monitor of a bridge pier.A three layer BP neural network is established to simulate the 3-dimention non-linear transformation with the use of neural network measure for network model selection and design. A new digital photographic technique is put forward combined camera self-calibration based on BP neural network with non-linear pixel coordinates correction of lens distortion. Meanwhile, the new technology is used to survey the tunnel contour. Compared with traditional methods, this method is more convenient and more adaptive.