| As a solid waste inevitably generated in urban construction,construction waste brings a series of environmental and social problems.It is seriously restrict the promotion of national circular economy and sustainable development,so it is particularly important to study the feature of construction waste and then be able to discover its spatial distribution quickly and accurately.The JL1-01 A remote sensing image developed by China.It has a high spatial resolution,while the revisit period just 3.3 days,which can reflect the slight change of the ground surface in quasi-real time,providing a new way to identify the spatial distribution and change detection of construction waste.However,the composition of construction waste and irregular accumulation boundaries lead to extremely complex spectral and geometric features on remote sensing images,which are extremely difficult to distinguish from the surrounding features,and the rapid changes and random distribution of accumulation sites bring great challenges to the rapid and high accuracy identification and detection of construction waste by remote sen sing,which has become a hot spot and difficult research area.In this paper,in view of the current situation of high cost and low efficiency of extracting construction waste information by traditional means,an in-depth study is carried out by combining the characteristics of JL1-01 A high resolution star data,the characteristic information of construction waste and the extraction method of construction waste in remote sensing images.Based on the analysis of the characteristics of JL1-01 A remote sensing image data,and based on the rich spectral,textural and geometric features and unique accumulation morphology of construction waste in the image as well as the surrounding physical environment,the real information in the near-infrared band of the GF-1 image is first introduced to compensate for the information in the simulated near-infrared band of JL1-01 A,and then the combined Relief F and J-M model algorithm to select the appropriate spectral,geometric and texture features,and then use a mul tifeature-based object-oriented image classification method to establish a construction waste knowledge rule model and extract the construction waste accumulation range to carry out,and finally use the CVA and SGD fusion approach to achieve rapid change detection of construction waste,while constructing an error confusion matrix to evaluate the accuracy of the experimental results and analyse the discussion.The main aspects of the study include the following.(1)In order to solve the problem that the near-infrared band of JL1-01 A remote sensing image is an analogue band,which may have certain influence on the recognition and information extraction of construction waste targets,etc.,and in view of its characteristics of this band,the image pre-processing method of adding real information of GF-1 near-infrared band to JL1-01 A remote sensing image is proposed to improve the band information of JL1-01 A satellite image.(2)To solve the problems of efficiency and accuracy of construction waste identification and extraction,the difficulty of complete and accurate identification and extraction of construction waste by single feature information,and the existence of greater subjectivity in manual selection of multiple features,the technical advantages of combining multiple features and object-oriented analysis methods are used to construct a knowledge rule classification model,adopt multi-scale segmentation techniques to obtain feature objects of different hierarchical levels,and use Relief F＿J-M,The combined model algorithm is used to achieve the preferential selection of spectral,texture and geometric features of construction waste,and then the optimized feature space and the object-oriented classification model based on rules are used to realize the o bject-oriented multifeature construction waste remote sensing information extraction based on JL1-01 A remote sensing images.(3)In order to verify the effectiveness of JL1-01 A remote sensing images for urban fine feature extraction,the necessity of introducing real NIR bands from other images and the rationality of constructing an optimized feature space for construction waste,a typical construction waste accumulation area in Daxing District of Beijing is extracted and the accuracy of construction waste classification before and after the improvement of band information is evaluated through field verification data and the establishment of confusion matrix.The accuracy of the extraction results is evaluated by field validation data and the establishment of confusion matrix.The results show that the construction waste feature selection and extraction method proposed in this paper can quickly achieve the feature preference of construction waste,and the overall accuracy of object-oriented classification reaches 87.6% and 91.3%,among which the accuracy of 87.6% is obtained by using the original JL1-01 A fusion image for object-oriented construction waste classification,and the accuracy increases to 91.3% after the improvement of real NIR band information.After improvement,the accuracy was increased to 91.3%.(4)On the basis of acquiring the spatial distribution location of construction waste,the construction waste change detection algorithm based on JL1-01 A remote sensing images is studied,and the threshold segmentation algorithm that can obtain the best construction waste change detection results is explored.The experimental results show that the CVA-SGD-based construction waste change detection algorithm can effectively detect the change of construction waste;each of the three threshold segmentation algorithms has its own advantages.Among them,the overall accuracy of the EM algorithm is better than the other two threshold segmentation algorithms,and the accuracy rate can reach over 90%.And increasing the adjustment factor can make the accuracy improve.When the adjustment factor is 0.2,the accuracy rate is 90.28%;when the adjustment factor is 0.8,the accuracy rate is 90.34%;the adaptive adjustment threshold method obtains the lowest false detection rate of 12.45%;and the global threshold method obtains the lowest missed detection rate of 11.56%.This research result is an important exploration of the use of remote sensing algorithms to detect changes in construction waste,which is of great sig nificance for the management department to realise the mastery of changes in construction waste,and has important promotion and application value for the application of domestic Jilin-1 satellite data to the extraction of fine feature information in cities. |
PDF Full Text Request