Albert K. Chong

Object-Based Classification of Ikonos Imagery for Mapping Large-Scale Vegetation Communities in Urban Areas

Effective assessment of biodiversity in cities requires detailed vegetation maps. To date, most remote sensing of urban vegetation has focused on thematically coarse land cover products. Detailed habitat maps are created by manual interpretation of aerial photographs, but this is time consuming and costly at large scale. To address this issue, we tested the effectiveness of object-based classifications that use automated image segmentation to extract meaningful ground features from imagery. We applied these techniques to very high resolution multispectral Ikonos images to produce vegetation community maps in Dunedin City, New Zealand. An Ikonos image was orthorectified and a multi-scale segmentation algorithm used to produce a hierarchical network of image objects. The upper level included four coarse strata: industrial/commercial (commercial buildings), residential (houses and backyard private gardens), vegetation (vegetation patches larger than 0.8/1ha), and water. We focused on the vegetation stratum that was segmented at more detailed level to extract and classify fifteen classes of vegetation communities. The first classification yielded a moderate overall classification accuracy (64%, κ = 0.52), which led us to consider a simplified classification with ten vegetation classes. The overall classification accuracy from the simplified classification was 77% with a κ value close to the excellent range (κ = 0.74). These results compared favourably with similar studies in other environments. We conclude that this approach does not provide maps as detailed as those produced by manually interpreting aerial photographs, but it can still extract ecologically significant classes. It is an efficient way to generate accurate and detailed maps in significantly shorter time. The final map accuracy could be improved by integrating segmentation, automated and manual classification in the mapping process, especially when considering important vegetation classes with limited spectral contrast.

Mapping Marine Habitats in Otago, Southern New Zealand

Colour aerial photographs and field surveys were used to assess the extent and dynamics of giant bladder kelp, Macrocystis pyrifera, and eelgrass, Zostera novazelandica, beds. A common framework was developed for acquiring, integrating and analysing the spatial data. The two biological systems present different remote sensing and management problems. These differences help demonstrate the need for specific consideration of natural biological cycles and local management issues when developing remote sensing projects for marine ecosystems. A good local understanding of the spatial and temporal dynamics of biological systems is essential to ensure that environmental indicators monitored using remote sensing are meaningful and that appropriate standards are set.

A combined stereo-photogrammetry and underwater-video system to study group composition of dolphins

Abstract  One reason for the paucity of knowledge of dolphin social structure is the difficulty of measuring individual dolphins. In Hector’s dolphins, Cephalorhynchus hectori, total body length is a function of age, and sex can be determined by individual colouration pattern. We developed a novel system combining stereo-photogrammetry and underwater-video to record dolphin group composition. The system consists of two downward-looking single-lens-reflex (SLR) cameras and a Hi8 video camera in an underwater housing mounted on a small boat. Bow-riding Hector’s dolphins were photographed and video-taped at close range in coastal waters around the South Island of New Zealand. Three-dimensional, stereoscopic measurements of the distance between the blowhole and the anterior margin of the dorsal fin (BH-DF) were calibrated by a suspended frame with reference points. Growth functions derived from measurements of 53 dead Hector’s dolphins (29 female : 24 male) provided the necessary reference data. For the analysis, the measurements were synchronised with corresponding underwater-video of the genital area. A total of 27 successful measurements (8 with corresponding sex) were obtained, showing how this new system promises to be potentially useful for cetacean studies.

Calibration and accuracy assessment of Leica ScanStation C10 terrestrial laser scanner

Requirement of high accuracy data in surveying applications has made calibration procedure a standard routine for all surveying instruments. This is due to the assumption that all observed data are impaired with errors. Thus, this routine is also applicable to terrestrial laser scanner (TLS) to make it available for surveying purposes. There are two calibration approaches: (1) component, and (2) system calibration. With the intention to specifically identify the errors and accuracy of the Leica ScanStation C10 scanner, this study investigates component calibration. Three components of calibration were performed to identify the constant, scale error, accuracy of angular measurement and the effect of angular resolution for distance measurement. The first calibration has been processed using closed least square solutions and has yielded the values of constant (1.2 mm) and scale error (1.000008879). Using variance ratio test (F-Test), angles observation (horizontal and vertical) for Leica C10 scanner and Leica TM5100A theodolite have shown significance difference. This is because the accuracy of both sensors are not similar and these differences are 0.01 and 0.0075o for horizontal and vertical measurements, respectively. Investigation on the resolution setting for Leica C10 scanner has highlighted the drawback of the tilt-and-turn target. Using the highest resolution, Leica Cyclone software only able to recognize the tilt-and-turn target up to 10 m distance compare to 200 m for the black and white target.

Virtual reality assessment of rugby lineout throw kinematics

The lineout throw in the sport of Rugby Union is a crucial component in determining whether a team wins possession of the ball. The accuracy and timing of the throw are both important factors in determining whether the opposition can intercept the ball, and these components are determined by the throwers technique and ability to identify the target. A virtual reality environment, NuView stereo-imaging device and a digital high-definition video camera were used to capture stereoscopic video footage of players during outdoor training. The video footage was presented to eight elite rugby players in life-size stereoscopic 3D as they attempted to perform a lineout as if in a real game. A custom-built laser device measured the accuracy of the virtual throws. In addition, a 12 camera motion analysis system tracked body movements, to help determine lineout throw variables associated with accurate throwing. In the trial of the virtual reality system, statistical and practically significant differences were found between accurate and inaccurate throws for the lineout throw variables; elbow separation (accurate 5.2% greater than inaccurate), front foot step (94.5 mm longer), trunk flexion (5°greater), and follow though (17°higher). These four lineout throw variables are recommended to coaches and players as important features to concentrate on when trying to enhance lineout throw accuracy.

TLS for generating multi-LOD of 3D building model

The popularity of Terrestrial Laser Scanners (TLS) to capture three dimensional (3D) objects has been used widely for various applications. Development in 3D models has also led people to visualize the environment in 3D. Visualization of objects in a city environment in 3D can be useful for many applications. However, different applications require different kind of 3D models. Since a building is an important object, CityGML has defined a standard for 3D building models at four different levels of detail (LOD). In this research, the advantages of TLS for capturing buildings and the modelling process of the point cloud can be explored. TLS will be used to capture all the building details to generate multi-LOD. This task, in previous works, involves usually the integration of several sensors. However, in this research, point cloud from TLS will be processed to generate the LOD3 model. LOD2 and LOD1 will then be generalized from the resulting LOD3 model. Result from this research is a guiding process to generate the multi-LOD of 3D building starting from LOD3 using TLS. Lastly, the visualization for multi-LOD model will also be shown.

Reliability of a high accuracy image-based system for 3D modelling of the medial longitudinal arch during gait

The Medial Longitudinal Arch (MLA) is the largest arch of the foot and is regarded as the most important foot arch in clinical foot assessments due to its influence on lower limb function, foot stability and foot pain. Each foot is classified as either high arched, low arched or normally arched depending on the structure of the MLA. There are currently a number of techniques that are used to classify the foot in a static state based on measurements of the MLA. These static measurements are then used to predict the behaviour of the foot arch in a dynamic state. However, it is easy to identify limitations with these techniques as the shape of the MLA in a static state cannot predict the behaviour of the MLA during dynamic activities. Therefore, the aim of this chapter is to introduce a high accuracy 3D modelling system that has been developed to map the shape of the MLA during gait using high definition video camcorders. The objectives of the study were hence: (1) to determine whether changes can be detected along the MLA for different weight bearings during gait, (2) to test the accuracy and reliability of the developed imaging system for creating dynamic 3D models of the foot arch and (3) to determine the quality and suitability of the 3D model. The results of the study show that changes can be detected along the MLA during gait with a level of accuracy of less than 0.4 mm when a 3D model of the foot is generated in PhotoModeler Scanner.

Night-time surveillance system for forensic 3D mapping

This paper discusses a research in the development of a night-time surveillance imaging system for forensic mapping, and the exploitation of the system for crime suspect identification. The motivation of this research was to develop a precision stereo-imaging system using an off-the-shelf imaging sensor for 3D mapping of crime suspects. The prototype surveillance imaging system consisted of a high definition video (HDV) camcorder, a 3D stereo adapter, an Infrared (IR) filter and an IR illuminator. Using the concept of stereo-photogrammetry for craniofacial anthropometric measurement, body height and distinctive body feature of a suspect can be measured accurately in 3D. This study focused on the development and calibration of the prototype for surveillance purposes. A practical system evaluation was carried out at an ATM booth where stereo-video footage of a crime scene was captured and processed. The results of the research showed that the prototype was capable of obtaining spatial data with 27.0 mm accuracy in faint-light conditions and within 3.0 m from the imaging system.

A CASE STUDY ON THE ESTABLISHMENT OF SHORELINE POSITION

Abstract In Malaysia, Mean High Water Mark (shoreline or riverbank) is a natural boundary for lands between the State and the coastal or estuary upland owners. The State. may also designate lands for coastal reserve or river reserve, which may cover 20 metres upland from the shoreline or the riverbank. In the State of Sarawak, Malaysia, an Act was passed in 1958 regarding the location of the coastal reserve or river reserve. However, shoreline or riverbank was not established accurately in some disputed native customary right land claims originating from the Act. Consequently, in 1983 seven evicted landowners brought suits against the Superintendent of Lands and Surveys and the government of Sarawak seeking compensation for their lands. In this paper the author highlights the methods used by two surveyors to determine the position of the shoreline from hydrographic charts, town plans and old aerial photographs. The court case shows that knowledge in the area of measurement error analysis; accurate photogrammetric technique and image processing are highly essential in the present-day court environment in which aerial photographic evidences are required.

Simulation of Interactive Cutting Tool for Craniofacial Osteotomy Planning

Engineers and scientists from many fields are using three-dimensional reconstruction for visualization and analysis of physical data. Recent advances in medical imaging and surgical techniques have made possible the correction of severe facial deformities and fractures. Surgical correction techniques often involve the direct manipulation - both relocation and surgical fracture - of the underlying facial bone. Craniofacial surgeons now a days use 3D models of human face to diagnose and analyze abnormalities and hence to plan their basic surgery procedures. To work with these 3D models, it is desirable to develop methods that allow one to interactively manipulate and alter the geometry in an intuitive and efficient manner. In the case of medical/craniofacial data, a baseline interactive task would be to simulate cutting, removal and realignment of tissue. In this paper, we propose a method to simulate a cutting tool using an oriented cube that allows the simulation of tool thickness as well as performing the desired cutting operations on three- dimensional models. This method can be employed interactively, allowing the user to perform the task in an intuitive and easy manner. The tool can be used for osteotomy planning and has been tested on models up to 500,000 polygons in size and the time for cutting varies from 15 ms for small objects (mandible) to 875 ms for large models (complete skull).