Research On Building Height Detection Of Remote Sensing Image

Remote sensing technology is a long-range goal of non-contact sensingtechnology. Remote sensing image contains a lot of wealth of knowledge data,extracting useful information on features from these images is a research hotspot ofthe discipline in recent years. High-resolution remote sensing image is often used inurban areas mapping work. With the continuous improvement of the spatial resolution,the the shadow from image is a category of factors which can not be ignored, Theshadow information has been long used to detect the height of buildings estimation byresearchers in the aviation surveying.Water bodies, lawns which is similar to shadow areas is often misidentified asgray shaded area in the traditional shadow detection method. Therefore, it hasgenerally significant limitations in application. Artificial neural networks, supportvector machines and other machine learning methods have earlier been used for dataclassification. But the traditional machine learning is difficult to select the parameters,and it lack of timeliness. when dealing with the large amounts of data. A new shadowdetection method based on extreme learning machine is proposed. Four statisticalfeatures, including energy, entropy, contrast and inverse difference moment, extractedfrom texture are used as the model input features to train learning machine model, theshadow area detection is realized. The problem of missed feature and repeatedthreshold calculation can be worked out by the proposed method. In addition, thedifficulty in selecting parameters of the neural network and support vector machinecan be solved, and the speed of shadow area detection is effectively improved. Thealgorithm also has better robustness and generalization, and the performance is alsobetter than that of threshold method, neural network and support vector machine.Building height extraction is a very important factor in the three-dimensionalreconstruction of the building. The relative geometry relationship between the shadowof a building, the sun and satellites is analyzed. A building height detection model isestablished in this paper. The shadow edge detection is detected by mathematicalmorphology method, and the length of the shadow is calculated by the corner detectionmethod. Finally the height of buildings from selected area is calculated and comparedwith the actual height. The effectiveness of the algorithm is verified.