Multiresolution Segmentation Method Research Articles

Extraction of Terraces on the Loess Plateau from High-Resolution DEMs and Imagery Utilizing Object-Based Image Analysis

Abstract: Terraces are typical artificial landforms on the Loess Plateau, with ecological functions in water and soil conservation, agricultural production, and biodiversity. Recording the spatial distribution of terraces is the basis of monitoring their extent and understanding their ecological effects. The current terrace extraction method mainly relies on high-resolution imagery, but its accuracy is limited due to vegetation coverage distorting the features of terraces in imagery. High-resolution topographic data reflecting the morphology of true terrace surfaces are needed. Terraces extraction on the Loess Plateau is challenging because of the complex terrain and diverse vegetation after the implementation of “vegetation recovery”. This study presents an automatic method of extracting terraces based on 1 m resolution digital elevation models (DEMs) and 0.3 m resolution Worldview-3 imagery as auxiliary information used for object-based image analysis (OBIA). A multi-resolution segmentation method was used where slope, positive and negative terrain index (PN), accumulative curvature slope (AC), and slope of slope (SOS) were determined as input layers for image segmentation by correlation analysis and Sheffield entropy method. The main classification features based on DEMs were chosen from the terrain features derived from terrain factors and texture features by gray-level co-occurrence matrix (GLCM) analysis; subsequently, these features were determined by the importance analysis on classification and regression tree (CART) analysis. Extraction rules based on DEMs were generated from the classification features with a total classification accuracy of 89.96%. The red band and near-infrared band of images were used to exclude construction land, which is easily confused with small-size terraces. As a result, the total classification accuracy was increased to 94%. The proposed method ensures comprehensive consideration of terrain, texture, shape, and spectrum characteristics, demonstrating huge potential in hilly-gully loess region with similarly complex terrain and diverse vegetation covers.

A Tool Assessing Optimal Multi-Scale Image Segmentation

Image segmentation to create representative objects by region growing image segmentation techniques such as multi resolution segmentation (MRS) is mostly done through interactive selection of scale parameters and is still a subject of great research interest in object-based image analysis. In this study, we developed an optimum scale parameter selector (OSPS) tool for objective determination of multiple optimal scales in an image by MRS using eCognition software. The ready to use OSPS tool consists of three modules and determines optimum scales in an image by combining intrasegment variance and intersegment spatial autocorrelation. The tool was tested using WorldView-2 and Resourcesat-2 LISS-IV Mx images having different spectral and spatial resolutions in two areas to find optimal objects for ground features such as water bodies, trees, buildings, road, agricultural fields and landslides. Quality of the objects created for these features using scale parameters obtained from the OSPS tool was evaluated quantitatively using segmentation goodness metrics. Results show that OSPS tool is able determine optimum scale parameters for creation of representative objects from high resolution satellite images by MRS method.

Integration of Ant Colony Optimization and Object-Based Analysis for LiDAR Data Classification

Light detection and ranging (LiDAR) data classification provides useful thematic maps for numerous geospatial applications. Several methods and algorithms have been proposed recently for LiDAR data classification. Most studies focused on object-based analysis because of its advantages over per-pixel-based methods. However, several issues, such as parameter optimization, attribute selection, and development of transferable rulesets, remain challenging in this topic. This study contributes to LiDAR data classification by developing an approach that integrates ant colony optimization (ACO) and rule-based classification. First, LiDAR-derived digital elevation and digital surface models were integrated with high-resolution orthophotos. Second, the processed raster was segmented with the multiresolution segmentation method. Subsequently, the parameters were optimized with a supervised technique based on fuzzy analysis. A total of 20 attributes were selected based on general knowledge on the study area and LiDAR data; the best subset containing 12 attributes was then selected via ACO. These attributes were utilized to develop rulesets through the use of a decision tree algorithm, and a thematic map was generated for the study area. Results revealed the robustness of the proposed method, which has an overall accuracy of ∼95% and a kappa coefficient of 0.94. The rule-based approach with all attributes and the k nearest neighbor (KNN) classification method were applied to validate the results of the proposed method. The overall accuracy of the rule-based method with all attributes was ∼88% (kappa = 0.82), whereas the KNN method had an overall accuracy of <70% and produced a poor thematic map. The selection of the ACO algorithm was justified through a comparison with three well-known feature selection methods. On the other hand, the transferability of the developed rules was evaluated by using a second LiDAR dataset at another study area. The overall accuracy and the kappa index for the second study area were 92% and 0.90, respectively. Overall, the findings indicate that the selection of a subset with significant attributes is important for accurate LiDAR data classification with object-based methods.

Delineating Individual Tree Crowns in an Uneven-Aged, Mixed Broadleaf Forest Using Multispectral Watershed Segmentation and Multiscale Fitting

Delineating individual tree crowns (ITCs) in high-spatial-resolution images can help to improve forest inventory and management. However, single-band watershed segmentation methods often fail to delineate broadleaf species, particularly in uneven-aged stands when a single-scale parameter is used to fit segments to reference crowns of different sizes. In this study, we present multispectral watershed segmentation and multiscale fitting method for ITC delineation, and the method involves two steps: 1) multispectral watershed segmentation to produce multiscale segmentation for subsequent fitting, which takes full advantage of boundary information contained in the spectral contrast of multiple bands; and 2) multiscale fitting to identify optimal parameters to best fit each ITC, rather than selecting a single parameter value based on its overall fit to the image as a whole. We evaluate the effectiveness of the proposed method using two multispectral images from a mixed broadleaf forest in Central Ontario, Canada. Our results show that multispectral watershed segmentation at multiple spatial scales produces ITC maps of higher quality than the commonly used multiresolution segmentation method. The automated multiscale fitting produces ITC maps of higher quality than the best single-scale segmentation.

Multi-scale segmentation of very high resolution remote sensing image based on gravitational field and optimized region merging

This paper proposes a multi-scale segmentation approach for high resolution remote sensing image (HRRSI) based on the gravitational field and region merging. In this approach, the HRRSI is firstly transformed into a gravitational field by incorporating the spatial and spectral information. Based on which, the attraction among neighboring pixels will cause travelling of each pixel. During the travelling, pixels with similar spectral information which are at the nearby location get grouped, which define an initial segmentation for the HRRSI which is often over-segmented in places. Then, an improved graph based region merging method is adopted to merge the over-segmented regions which yield the multi-scale segmentation results. To evaluate the proposed approach, we conduct extensive experiments on three HRRSIs from different sensors and the obtained results are compared with those of eCognition's multi-resolution segmentation method. The experimental results show that the proposed approach reduces much more over-segmentation problem and produces more accurate image segmentation.

UAS photogrammetry and object-based image analysis (GEOBIA): erosion monitoring at the Kazár badland, Hungary

A remarkable badland valley is situated near Kazár, NE-Hungary, where rhyolite tuff outcrops as greyishwhite cliffs and white barren patches. The landform is shaped by gully and rill erosion processes. Weperformed a preliminary state UAS survey and created a digital surface model and ortophotograph. Theflight was operated with manual control in order to perform a more optimal coverage of the aerial images.The overhanging forests induced overexposed photographs due to the higher contrast with the baretuff surface. The multiresolution segmentation method allowed us to classify the ortophotograph andseparate the tuff surface and the vegetation. The applied methods and final datasets in combination withthe subsequent surveys will be used for detecting the recent erosional processes of the Kazár badland

The Application of GF-1 Imagery to Detect Ships on the Yangtze River

Various satellite data are currently used to detect ships on the sea surface. However, no study on the use of Gaofen-1 (GF-1) data to monitor ships on the surface of inland rivers has been reported. Therefore, we proposed a method to extract inland river-surface ships from GF-1 imagery. The Normalized Differential Water Index was calculated to enhance the contrast between water and non-water areas after the preprocessing procedure. The multi-resolution segmentation method and object-oriented classification rule sets were used to detect the ships in the image. Results show that most of the ships, whose length-to-width ratio ranges from 3.0 to 7.2, could be identified correctly regardless of their size. The results also indicate that detecting ships on inland rivers using GF-1 imagery is feasible.

Empirical comparison of color normalization methods for epithelial-stromal classification in H and E images

Context: Color normalization techniques for histology have not been empirically tested for their utility for computational pathology pipelines. Aims: We compared two contemporary techniques for achieving a common intermediate goal – epithelial-stromal classification. Settings and Design: Expert-annotated regions of epithelium and stroma were treated as ground truth for comparing classifiers on original and color-normalized images. Materials and Methods: Epithelial and stromal regions were annotated on thirty diverse-appearing H and E stained prostate cancer tissue microarray cores. Corresponding sets of thirty images each were generated using the two color normalization techniques. Color metrics were compared for original and color-normalized images. Separate epithelial-stromal classifiers were trained and compared on test images. Main analyses were conducted using a multiresolution segmentation (MRS) approach; comparative analyses using two other classification approaches (convolutional neural network [CNN], Wndchrm) were also performed. Statistical Analysis: For the main MRS method, which relied on classification of super-pixels, the number of variables used was reduced using backward elimination without compromising accuracy, and test - area under the curves (AUCs) were compared for original and normalized images. For CNN and Wndchrm, pixel classification test-AUCs were compared. Results: Khan method reduced color saturation while Vahadane reduced hue variance. Super-pixel-level test-AUC for MRS was 0.010–0.025 (95% confidence interval limits ± 0.004) higher for the two normalized image sets compared to the original in the 10–80 variable range. Improvement in pixel classification accuracy was also observed for CNN and Wndchrm for color-normalized images. Conclusions: Color normalization can give a small incremental benefit when a super-pixel-based classification method is used with features that perform implicit color normalization while the gain is higher for patch-based classification methods for classifying epithelium versus stroma.

Multiresolution saliency map based object segmentation

Salient objects’ detection and segmentation are gaining increasing research interest in recent years. A saliency map can be obtained from different models presented in previous studies. Based on this saliency map, the most salient region (MSR) in an image can be extracted. This MSR, generally a rectangle, can be used as the initial parameters for object segmentation algorithms. However, to our knowledge, all of those saliency maps are represented in a unitary resolution although some models have even introduced multiscale principles in the calculation process. Furthermore, some segmentation methods, such as the well-known GrabCut algorithm, need more iteration time or additional interactions to get more precise results without predefined pixel types. A concept of a multiresolution saliency map is introduced. This saliency map is provided in a multiresolution format, which naturally follows the principle of the human visual mechanism. Moreover, the points in this map can be utilized to initialize parameters for GrabCut segmentation by labeling the feature pixels automatically. Both the computing speed and segmentation precision are evaluated. The results imply that this multiresolution saliency map-based object segmentation method is simple and efficient.

Fusion of multiple features to produce a segmentation algorithm for remote sensing images

This letter presents an edge direction adaptive watershed segmentation method for remote sensing images. First, the maximum gradient value among different directions is chosen as the single band gradient value of the pixel, and a compound gradient value is then calculated based on the gradient value in each band. Second, the marker-based watershed segmentation is implemented to produce initial over-segmentation result to avoid under-segmentation. Finally, the adjacent objects with high similarity values are merged to reduce over-segmentation, which improves segmentation accuracy. The performance of the proposed method is validated on two satellite images. Experimental results show that, compared with the multi-resolution segmentation method embedded in the eCognition software and the traditional multi-band watershed segmentation method, the proposed method can decrease over-/under-segmentation and thus produce satisfactory segmentation results.

Investigation of Optimal Segmentation Color Space of Bayer True Color Images with Multi-Objective Optimization Methods

For the selection of the optimal segmentation space of Bayer true color unmanned aerial vehicle image, this paper introduces multi-objectives constraints optimization to solve the inconsistency of multiple indicators. First, the Bayer color images were converted to YIQ(Luninance, Inphase, Quadrature), YCbCr(Luninance, Blue-difference, Red-difference), I1I2I3 (Three linear transformed color-opponent dimensions), HSI(Hue, Saturation, Intensity), Nrgb(Normalized Red, Green, Blue) and CIE(L*a*b*) (Comission Internationale de l′Eclairage, L*a*b* for Lightness and two color-opponent dimensions)color space, then the transformed images were segmented with multi-resolution segmentation method. By introducing the multi-objective constraint function, three parameters such as the topology index, geometric index and spectral area matching index were synthetically considered to determine the optimal segmentation scale. Based on that, the multi-objective constraint function was built to comprehensively analyze the result of segmentation, so as to find out the optimal color space for a certain type of building. And then the global optimum color space appropriate for all kinds of buildings can be gained through the comprehensive analysis of the F value of different types of buildings. Finally a series of images of different acquisition conditions and ground features were selected to conduct the test. The result shows that the optimal segmentation color spaces of different types of buildings vary a little. For cottage the I1I2I3 space can get the excellent object areas that reflect the real edge of the ground features, while the YCbCr space has some advantages on the segmentation of tile-building. Overall, only I1I2I3 color space has better integrated segmentation result for all buildings, and it is considered to be the best color space suitable for segmentation.

Segmentation quality evaluation using region-based precision and recall measures for remote sensing images

Segmentation of remote sensing images is a critical step in geographic object-based image analysis. Evaluating the performance of segmentation algorithms is essential to identify effective segmentation methods and optimize their parameters. In this study, we propose region-based precision and recall measures and use them to compare two image partitions for the purpose of evaluating segmentation quality. The two measures are calculated based on region overlapping and presented as a point or a curve in a precision–recall space, which can indicate segmentation quality in both geometric and arithmetic respects. Furthermore, the precision and recall measures are combined by using four different methods. We examine and compare the effectiveness of the combined indicators through geometric illustration, in an effort to reveal segmentation quality clearly and capture the trade-off between the two measures. In the experiments, we adopted the multiresolution segmentation (MRS) method for evaluation. The proposed measures are compared with four existing discrepancy measures to further confirm their capabilities. Finally, we suggest using a combination of the region-based precision–recall curve and the F-measure for supervised segmentation evaluation.

Biomass estimation with high resolution satellite images: A case study of Quercus rotundifolia

Forest biomass has had a growing importance in the world economy as a global strategic reserve, due to applications in bioenergy, bioproduct development and issues related to reducing greenhouse gas emissions. Current techniques used for forest inventory are usually time consuming and expensive. Thus, there is an urgent need to develop reliable, low cost methods that can be used for forest biomass estimation and monitoring. This study uses new techniques to process high spatial resolution satellite images (0.70m) in order to assess and monitor forest biomass. Multi-resolution segmentation method and object oriented classification are used to obtain the area of tree canopy horizontal projection for Quercus rotundifolia. Forest inventory allows for calculation of tree and canopy horizontal projection and biomass, the latter with allometric functions. The two data sets are used to develop linear functions to assess above ground biomass, with crown horizontal projection as an independent variable. The functions for the cumulative values, both for inventory and satellite data, for a prediction error equal or smaller than the Portuguese national forest inventory (7%), correspond to stand areas of 0.5ha, which include most of the Q.rotundifolia stands.

Hybrid region merging method for segmentation of high-resolution remote sensing images

Image segmentation remains a challenging problem for object-based image analysis. In this paper, a hybrid region merging (HRM) method is proposed to segment high-resolution remote sensing images. HRM integrates the advantages of global-oriented and local-oriented region merging strategies into a unified framework. The globally most-similar pair of regions is used to determine the starting point of a growing region, which provides an elegant way to avoid the problem of starting point assignment and to enhance the optimization ability for local-oriented region merging. During the region growing procedure, the merging iterations are constrained within the local vicinity, so that the segmentation is accelerated and can reflect the local context, as compared with the global-oriented method. A set of high-resolution remote sensing images is used to test the effectiveness of the HRM method, and three region-based remote sensing image segmentation methods are adopted for comparison, including the hierarchical stepwise optimization (HSWO) method, the local-mutual best region merging (LMM) method, and the multiresolution segmentation (MRS) method embedded in eCognition Developer software. Both the supervised evaluation and visual assessment show that HRM performs better than HSWO and LMM by combining both their advantages. The segmentation results of HRM and MRS are visually comparable, but HRM can describe objects as single regions better than MRS, and the supervised and unsupervised evaluation results further prove the superiority of HRM.

Segmentation of High Spatial Resolution Remote Sensing Imagery Based on Hard-Boundary Constraint and Two-Stage Merging

This paper proposes a novel two-stage method for remote sensing image segmentation. First, initial small segments, also called subobject primitives (sub-OPs), are obtained using edge-constrained watershed segmentation and edge allocation. These segments are gradually merged into a larger segment until the edge-controlled limits are reached, thereby creating the initial OPs. In this stage, a concept of hard-boundary ratio is proposed to control the merge effectively. Second, nonconstrained merging is conducted on the OPs, which results in final segmentation. In addition, a repeatable pairwise segment-merging scheme is utilized. This scheme improves method efficiency and accuracy. Comprehensive experiments comparing this new method with the multiresolution segmentation method of eCognition were conducted. Results show that this new method has the following advantages: 1) higher segmentation accuracy and OP boundary precision and 2) less dependence on the scale parameter.

Object-Based Land Cover Classification for ALOS Image Combining TM Spectral

Abstract. Land cover classification for high spatial resolution remote sensing images becomes a challenging work. The high spatial resolution remote sensing images have more spatial information. The low or medium resolution remote sensing images have more spectral information. In order to improve the accuracy of high spatial resolution remote sensing image classification, additional information should be incorporated into the classification process of high spatial resolution remote sensing image. This paper proposed a method of object-based land cover classification for high spatial resolution ALOS images combining the spectral information of TM images. First, the high spatial resolution ALOS panchromatic image was segmented by multi-resolution segmentation method. Second, the spectral features of segmented regions were extracted from multi-spectral ALOS image and TM image by spatial mapping mechanism. Third, the regions were classified by SVM classifier. Experimental results show that the classification method for high spatial resolution remote sensing images combining the TM spectral information based on the spatial mapping mechanism can make use of the spectral information both in high and low spatial resolution remote sensing images and improve classification accuracy.

Preliminary results for a semi-automated quantification of site effects using geomorphometry and ASTER satellite data for Mozambique, Pakistan and Turkey

Estimation of the degree of local seismic wave amplification (site effects) requires precise information about the local site conditions. In many regions of the world, local geologic information is either sparse or is not readily available. Because of this, seismic hazard maps for countries such as Mozambique, Pakistan and Turkey are developed without consideration of site factors and, therefore, do not provide a complete assessment of future hazards. Where local geologic information is available, details on the traditional maps often lack the precision (better than 1:10,000 scale) or the level of information required for modern seismic microzonation requirements. We use high-resolution (1:50,000) satellite imagery and newly developed image analysis methods to begin addressing this problem. Our imagery, consisting of optical data and digital elevation models (DEMs), is recorded from the ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) sensor system. We apply a semi-automated, object-oriented, multi-resolution feature segmentation method to identify and extract local terrain features. Then we classify the terrain types into mountain, piedmont and basin units using geomorphometry (topographic slope) as our parameter. Next, on the basis of the site classification schemes from the Wills and Silva (1998) study and the Wills et al (2000) and Wills and Clahan (2006) maps of California, we assign the local terrain units with Vs30 (the average seismic shear-wave velocity through the upper 30m of the subsurface) ranges for selected regions in Mozambique, Pakistan and Turkey. We find that the applicability of our site class assignments in each region is a good first-approximation for quantifying local site conditions and that additional work, such as the verification of the terrain’s compositional rigidity, is needed.

Effective multiresolution arc segmentation: algorithms and performance evaluation

Arc segmentation plays an important role in the process of graphics recognition from scanned images. The GREC arc segmentation contest shows there is a lot of room for improvement in this area. This paper proposes a multiresolution arc segmentation method based on our previous seeded circular tracking algorithm which largely depends on the OOPSV model. The newly-introduced multiresolution paradigm can handle arcs/circles with large radii well. We describe new approaches for arc seed detection, arc localization, and arc verification, making the proposed method self-contained and more efficient. Moreover, this paper also brings major improvement to the dynamic adjustment algorithm of circular tracking to make it more robust. A systematic performance evaluation of the proposed method has been conducted using the third-party evaluation tool and test images obtained from the GREC arc segmentation contests. The overall performance over various arc angles, arc lengths, line thickness, noises, arc-arc intersections, and arc-line intersections has been measured. The experimental results and time complexity analyses on real scanned images are also reported and compared with other approaches. The evaluation result demonstrates the stable performance and the significant improvement on processing large arcs/circles of the MAS method.

A novel multiresolution fuzzy segmentation method on MR image

Multiresolution-based magnetic resonance (MR) image segmentation has attracted attention for its ability to capture rich information across scales compared with the conventional segmentation methods. In this paper, a new scale-space-based segmentation model is presented, where both the intra-scale and inter-scale properties are considered and formulated as two fuzzy energy functions. Meanwhile, a control parameter is introduced to adjust the contribution of the similarity character across scales and the clustering character within the scale. By minimizing the combined inter/intra energy function, the multiresolution fuzzy segmentation algorithm is derived. Then the coarse to fine leading segmentation is performed automatically and iteratively on a set of multiresolution images. The validity of the proposed algorithm is demonstrated by the test image and pathological MR images. Experiments show that by this approach the segmentation results, especially in the tumor area delineation, are more precise than those of the conventional fuzzy segmentation methods.

Unconstrained handwritten numeral recognition based on radial basis competitive and cooperative networks with spatio-temporal feature representation

This paper presents a new approach to representation and recognition of handwritten numerals. The approach first transforms a two-dimensional (2-D) spatial representation of a numeral into a three-dimensional (3-D) spatio-temporal representation by identifying the tracing sequence based on a set of heuristic rules acting as transformation operators. A multiresolution critical-point segmentation method is then proposed to extract local feature points, at varying degrees of scale and coarseness. A new neural network architecture, referred to as radial-basis competitive and cooperative network (RCCN), is presented especially for handwritten numeral recognition. RCCN is a globally competitive and locally cooperative network with the capability of self-organizing hidden units to progressively achieve desired network performance, and functions as a universal approximator of arbitrary input-output mappings. Three types of RCCNs are explored: input-space RCCN (IRCCN), output-space RCCN (ORCCN), and bidirectional RCCN (BRCCN). Experiments against handwritten zip code numerals acquired by the U.S. Postal Service indicated that the proposed method is robust in terms of variations, deformations, transformations, and corruption, achieving about 97% recognition rate.