Technology Research Of Forest Landscape Remote Sensing Image Classification Based On The Hadoop Platform

At present, with the changes of forest management idea, each group has different emphasis on the benefits of forest management, and the management mode of forest management is becoming more and more complex. Therefore, the difficulty of sustainable forest management is greatly increased. Due to the limitations of the complexity of the forest landscape, remote sensing and classification algorithm, there are a variety of uncertainties by using remote sensing data for forest planning in the whole process. So how to improve the accuracy and efficiency has become a severe test. Remote sensing image classification is an important method for the forest landscape planning, and the application of remote sensing image classification technique includes image acquisition and pre-processing, image classification algorithm selection, etc. The information acquisition and the target recognition are the core technology of remote sensing applications. In this paper, from the view of image acquisition and algorithm performance, then improve the accuracy of access to information.This article is mainly including several items:(1) In image acquisition, the sensor can locate the image to 0.01 pixels (located in the sub pixel level), which can meet the high precision positioning of moving image. Ultimately improve the accuracy and defense capabilities of the satellite information acquisition. However, due to its own characteristics, it is very easy to appear the key failure problem, which makes its life expectancy greatly reduced. This becomes one of the bottlenecks in the market. In this paper, the TRIZ theory is used to optimize this problem, and then reduce the production cost of the device. It is feasible to use the sensor to improve the accuracy of the information acquisition.(2) The sub-pixel mapping method based on simulated annealing algorithm is introduced into the forest landscape classification research, by optimizing the spatial distribution of sub pixel, and makes sure the position of each mixed pixel points finally. That can improve the classification accuracy of forest landscape. Considering the large-scale data set problem of remote sensing image, optimizes simulated annealing algorithm by means of a series of parallel optimization strategy(independent and cooperative search, with or without communication and local and global optimization), and then improves the performance of the algorithm.(3) The serial and parallel algorithms are applied in the forest landscape classification, and researchers calculate the time complexity according to the performance of different algorithms. Then, contrast the experimental results of the parallel algorithm and serial algorithm. After analysis, the parallel independent searches converge much faster than the sequential algorithm. The researchers examine the performance of parallel algorithms on bench-marking tests elaborated by Solomon, both obtained higher efficiency. In theory and experiment, that shows the parallel strategy can be efficient for the forest landscape classification, which is effectively carried out on the forest landscape planning.