深度学习方法应用于深圳市多调合一工作的思考

Reflection on the application of deep learning method in “multiple-investigation-coordination” in Shenzhen

多调合一是实现调查成果数据融合,从源头上解决数据多样、数据冲突等问题的有效手段。该项工作的完成需要结合地区的实际情况,建立统一的调查监测评价体系,协同开展基础调查和专项调查工作。因此,首先需要在空间上保持基础调查单元划分的有效性,才能保证专项调查单元能够通过基础调查单元进行归并及获取;其次,需要在基础调查和专题调查中保证类别判定的准确性。本文结合深度学习方法在目前影像分类和目标识别等领域中的应用进展,对其在深圳市多调合一工作中的应用前景进行了思考,从减少人力物力的消耗、提高类别判定精度等多个角度进行了分析。随着深度学习方法的不断完善,未来其在自然资源调查领域的应用效果将会得到不断提升。

Multiple-investigation-coordination is an effective method to achieve data fusion from varies kinds of investigation activities which solves the problem of heterogeneity and conflict of data source. The foundation of the work requires not only unified investigation and estimation system but also synchronously pushing forward the basic investigation and special investigation based on the local circumstances. As a result,to make sure the correctness of acquisition of investigate unit of special investigation,the first priority is to keep the validity of the basic unit division. Moreover,the correctness of classification in the both investigating activities should be guaranteed. In the paper,we consider the application prospect of deep learning method of image classification and object recognition in the"Multiple-investigation-coordination"works in Shenzhen in order to reduce the human labor compared with traditional outdoor investigating tasks and analyze how to improve the accuracy of the classification result. With the continuous improvement of deep learning method,the effectiveness of such method applied in the natural resource investigation activities will enhance constantly.