基于图像识别山地城市绿色空间景观生态破损区域监测技术.docVIP

基于图像识别山地城市绿色空间景观生态破损区域监测技术 　　摘 要： 传统区域监测方法不能根据山地城市绿色空间景观生态破损区域的具体图像情况进行实时监测。为了有效解决此问题，建立新型基于图像识别的山地城市绿色空间景观生态破损区域实时监测模型。通过山地城市绿色空间景观生态破损区域特征判断、基于特征判断的识别方法确定、卷积神经网络，完成山地城市绿色空间景观生态破损区域的图像识别。通过实时目标检测、确定实时监测区域、透视变换，完成基于图像识别区域实时监测模型的搭建。设计对比实验结果表明，新型区域监测模型与传统模型相比，充分提升了山地城市绿色空间景观生态破损区域识别图像的清晰度，并可对区域内景观进行实时监测。 　　关键词： 图像识别; 山地城市; 绿色空间景观; 生态破损区域; 区域监测; 特征判断; 卷积神经网络; 透视变换 　　中图分类号： TN911.73?34; TP393 文献标识码： A 文章编号： 1004?373X（2018）12?0067?04 　　Abstract： As the real?time monitoring of ecological damage area cannot be performed according to its specific images for green space landscape of mountainous cities in the traditional area monitoring method， a new ecological damage area real?time monitoring model for green space landscape of mountainous cities is established based on image recognition. Ecological damage area image recognition for green space landscape of mountainous cities is accomplished by means of ecological damage area feature judgment for green space landscape of mountainous cities， recognition method determination based on feature judgment， and convolutional neural network. Establishment of image recognition based area real?time monitoring model is accomplished by means of real?time target monitoring， determination of real?time monitoring area， and perspective transformation. A comparison experiment was designed. The results show that in comparison with the traditional model， the new area monitoring model can fully improve the definition of ecological damage area recognition images for green space landscape of mountainous cities， and perform real?time monitoring of the landscape in the area. 　　Keywords： image recognition; mountainous city; green space landscape; ecological damage area; area monitoring; feature judgment; convolutional neural network; perspective transformation 　　0 引 言 　　传统区域监测模型利用人工智能理论，通过搭建人工神经网络、深层神经网络等方式，完成对区域内景观的监测。这种方式虽然全面体现了神经网络算法的重要意义，但随着科学技术手段的发展，传统模型不能对区域内景观进行实时监测的弊端开始逐渐显露出来。对于山地城市绿色空间景观生态破损区域来说，传统区域检测模型可以通过确定区域大小、简单描画区域特征的方式，生成基础的识别图