Manchun Li

A systematic comparison of different object-based classification techniques using high spatial resolution imagery in agricultural environments

Abstract Geographic Object-Based Image Analysis (GEOBIA) is becoming more prevalent in remote sensing classification, especially for high-resolution imagery. Many supervised classification approaches are applied to objects rather than pixels, and several studies have been conducted to evaluate the performance of such supervised classification techniques in GEOBIA. However, these studies did not systematically investigate all relevant factors affecting the classification (segmentation scale, training set size, feature selection and mixed objects). In this study, statistical methods and visual inspection were used to compare these factors systematically in two agricultural case studies in China. The results indicate that Random Forest (RF) and Support Vector Machines (SVM) are highly suitable for GEOBIA classifications in agricultural areas and confirm the expected general tendency, namely that the overall accuracies decline with increasing segmentation scale. All other investigated methods except for RF and SVM are more prone to obtain a lower accuracy due to the broken objects at fine scales. In contrast to some previous studies, the RF classifiers yielded the best results and the k-nearest neighbor classifier were the worst results, in most cases. Likewise, the RF and Decision Tree classifiers are the most robust with or without feature selection. The results of training sample analyses indicated that the RF and adaboost. M1 possess a superior generalization capability, except when dealing with small training sample sizes. Furthermore, the classification accuracies were directly related to the homogeneity/heterogeneity of the segmented objects for all classifiers. Finally, it was suggested that RF should be considered in most cases for agricultural mapping.

Extraction of Urban Power Lines from Vehicle-Borne LiDAR Data

Airborne LiDAR has been traditionally used for power line cruising. Nevertheless, data acquisition with airborne LiDAR is constrained by the complex environments in urban areas as well as the multiple parallel line structures on the same power line tower, which means it is not directly applicable to the extraction of urban power lines. Vehicle-borne LiDAR system has its advantages upon airborne LiDAR and this paper tries to utilize vehicle-borne LiDAR data for the extraction of urban power lines. First, power line points are extracted using a voxel-based hierarchical method in which geometric features of each voxel are calculated. Then, a bottom-up method for filtering the power lines belonging to each power line is proposed. The initial clustering and clustering recovery procedures are conducted iteratively to identify each power line. The final experiment demonstrates the high precision of this technique.

Semi-Automatic Registration of Airborne and Terrestrial Laser Scanning Data Using Building Corner Matching with Boundaries as Reliability Check

Data registration is a prerequisite for the integration of multi-platform laser scanning in various applications. A new approach is proposed for the semi-automatic registration of airborne and terrestrial laser scanning data with buildings without eaves. Firstly, an automatic calculation procedure for thresholds in density of projected points (DoPP) method is introduced to extract boundary segments from terrestrial laser scanning data. A new algorithm, using a self-extending procedure, is developed to recover the extracted boundary segments, which then intersect to form the corners of buildings. The building corners extracted from airborne and terrestrial laser scanning are reliably matched through an automatic iterative process in which boundaries from two datasets are compared for the reliability check. The experimental results illustrate that the proposed approach provides both high reliability and high geometric accuracy (average error of 0.44 m/0.15 m in horizontal/vertical direction for corresponding building corners) for the final registration of airborne laser scanning (ALS) and tripod mounted terrestrial laser scanning (TLS) data.

A Caution on the Use of Surface Digital Elevation Models to Simulate Supraglacial Hydrology of the Greenland Ice Sheet

Digital elevation models (DEMs) of ice surface topography are often used for hydrologic analysis of the Greenland ice sheet (GrIS), but their suitability for this purpose has received little quantitative assessment. We compare remotely sensed maps of supraglacial lakes, rivers, and moulins with their DEM-modeled counterparts, using two moderate-resolution DEMs (SPIRIT DEM and ASTER GDEM) for a ~24 000 km2 area of the southwestern GrIS. We find that modeled hydrological features are critically sensitive to selection of a depression area threshold (a user-specified parameter used to fill noise and/or true topographic depressions in the ice surface DEM), with small depression area thresholds over-predicting observed supraglacial lake abundance and large thresholds under-predicting lake abundance. Few remotely sensed moulins are identified in either DEM, even if a small depression area threshold is used. A standard practice of filling all DEM depressions yields modeled surface flow paths that broadly match remotely sensed supraglacial river networks, but are far less fragmented than reality (owing to moulin capture), raising into question the realism of this standard practice. In sum, moderate-resolution DEMs do hold value for simulating broad-scale hydrography of the GrIS surface, but are critically sensitive to choice of the filling threshold, and insensitive to moulins which also influence supraglacial drainage pattern. Our preliminary analysis suggests using a depression area threshold of 0.1-0.2 km2 for lakes, advises against using DEMs to predict moulin locations, and urges caution when using 100% DEM filling to model flow paths of supraglacial rivers.

Evaluation of Feature Selection Methods for Object-Based Land Cover Mapping of Unmanned Aerial Vehicle Imagery Using Random Forest and Support Vector Machine Classifiers

The increased feature space available in object-based classification environments (e.g., extended spectral feature sets per object, shape properties, or textural features) has a high potential of improving classifications. However, the availability of a large number of derived features per segmented object can also lead to a time-consuming and subjective process of optimizing the feature subset. The objectives of this study are to evaluate the effect of the advanced feature selection methods of popular supervised classifiers (Support Vector Machines (SVM) and Random Forest (RF)) for the example of object-based mapping of an agricultural area using Unmanned Aerial Vehicle (UAV) imagery, in order to optimize their usage for object-based agriculture pattern recognition tasks. In this study, several advanced feature selection methods were divided into both types of classifiers (SVM and RF) to conduct further evaluations using five feature-importance-evaluation methods and three feature-subset-evaluation methods. A visualization method was used to measure the change pattern of mean classification accuracy with the increase of features used, and a two-tailed t-test was used to determine the difference between two population means for both repeated ten classification accuracies. This study mainly contribute to the uncertainty analysis of feature selection for object-based classification instead of the per-pixel method. The results highlight that the RF classifier is relatively insensitive to the number of input features, even for a small training set size, whereby a negative impact of feature set size on the classification accuracy of the SVM classifier was observed. Overall, the SVM Recursive Feature Elimination (SVM-RFE) seems to be an appropriate method for both groups of classifiers, while the Correlation-based Feature Selection (CFS) is the best feature-subset-evaluation method. Most importantly, this study verified that feature selection for both classifiers is crucial for the evolving field of Object-based Image Analysis (OBIA): It is highly advisable for feature selection to be performed before object-based classification, even though an adverse impact could sometimes be observed from the wrapper methods.

Multiscale Grid Method for Detection and Reconstruction of Building Roofs from Airborne LiDAR Data

This study proposes a multiscale grid method to detect and reconstruct building roofs from airborne LiDAR data. The method interpolates unorganized LiDAR point cloud into two sets of grids with different spatial scales. In the large-scale grid, building seed regions are obtained, including detection of initial building seed regions and refinement of building seed regions. In the small-scale grid, to detect the detailed features of building roofs with complicated top structures, a high-resolution depth image is generated by a new iterative morphological interpolation using gradually increasing scales, and then segmented by using a full λ-schedule algorithm. Based on the building seed regions, detailed roof features are detected for each building and 3-D building roof models are then reconstructed according to the elevation of these features. Experiments are analyzed from several aspects: the correctness and completeness, the elevation accuracy of building roof models, and the influence of elevation to 3-D roof reconstruction. The experimental results demonstrate promising correctness, completeness, and elevation accuracy, with a satisfactory 3-D building roof models. The strategy of hierarchical spatial scale (from large scale to small scale) obtains the complementary advantage between technical applicability in a large urban environment and high quality in 3-D reconstruction of building roofs with fine details.

Using unmanned aerial vehicle for remote sensing application

With the availability and development of multisensor, multitemporal, multiresolution, and multifrequency image data from operational Earth observation satellites, more limitations have also appeared in remote sensing image applications, and the most important one is the cost limit. To overcome the limitations, unmanned aerial vehicle (UAV) technology is developped for remote sensing application. It forms a rapidly developing area of research in remote sensing. This review paper describes and explains data acquisition and processing technologies for imageries from UAV, as well as summarizes the current application research with UAV. Then, it presents some problems and prospective for UAV development. It is our hope that this survey will provide guidelines for future applications of UAV and possible areas.

Study on shadow detection method on high resolution remote sensing image based on HIS space transformation and NDVI index

This paper studied on shadow detection method on high resolution remote sensing image based on HIS space transformation and NDVI index. Firstly, the high resolution remote sensing image was transformed to HSI space, and the shadow of image could be detected with the characteristics of the low brightness and the high saturation in the area of shadow. Then according to the NDVI index of image, some of the ground objects, such as dark green grass, could be eliminated, and the shadow may be detected more efficaciously and accurately.

Seasonal Pattern of Tidal-Flat Topography along the Jiangsu Middle Coast, China, Using HJ-1 Optical Images

Seasonal topographical changes in the intertidal zone are of high interest in many parts of the world. Due to existing challenges in mapping this dynamics, little is known about the seasonal pattern of tidal flats. This research aims to fill the knowledge gap by using optical images from the Chinese HJ-1 satellites constellation. A case study was conducted at the Jiangsu middle coast, a typical tidal flat region in China. Firstly, 455 optical images from the HJ-1 Satellites were collected to construct seasonal tidal flat Digital Elevation Models (DEMs). Next, synchronous Light Detection and Ranging (LiDAR) DEM and ground survey data were collected to validate the accuracy of the seasonal tidal flat DEMs. Finally, seasonal pattern of the tidal flats was demonstrated based on a time series of DEM volume comparison. The results show: (1) the HJ-1 images are qualified data source to produce satisfactory tidal flat DEMs with high spatio-temporal resolution and acceptable vertical accuracy. (2) In general, there are apparent erosion-and-deposition cycles in the Jiangsu middle coast, with deposition during the Winter and erosion during the Summer. Furthermore, the derived seasonal patterns differ notably among the five major sandbanks.

Cultivated land information extraction from high-resolution unmanned aerial vehicle imagery data

Abstract The development of precision agriculture demands high accuracy and efficiency of cultivated land information extraction. Simultaneously, unmanned aerial vehicles (UAVs) have been increasingly used for natural resource applications in recent years as a result of their greater availability, the miniaturization of sensors, and the ability to deploy UAVs relatively quickly and repeatedly at low altitudes. We examine the potential of utilizing a small UAV for the characterization, assessment, and monitoring of cultivated land. Because most UAV images lack spectral information, we propose a novel cultivated land information extraction method based on a triangulation for cultivated land information extraction (TCLE) method. Thus, the information on more spatial properties of a region is incorporated into the classification process. The TCLE comprises three main steps: image segmentation, triangulation construction, and triangulation clustering using AUTOCLUST. Experiments were conducted on three UAV images in Deyang, China, using TCLE and eCognition for cultivated land information extraction (ECLE). Experimental results show that TCLE, which does not require training samples and has a much higher level of automation, can obtain accuracies equivalent to ECLE. Comparing with ECLE, TCLE also extracts coherent cultivated land with much less noise. As such, cultivated land information extraction based on high-resolution UAV images can be effectively and efficiently conducted using the proposed method.