Olga Duran

State of the art in sensor technologies for sewer inspection

This paper reviews the state of the art in sensors and automated inspection devices for enhanced sewer inspection. Efficiency, safety, environmental, and legislative concerns have made inspection and assessment of communal sewers a central issue to water and sewerage companies. Nowadays, the standard sewer inspection system is based on a wheeled platform on which a closed circuit television (CCTV) camera is mounted. One of the disadvantages of camera inspection systems is that they can only detect a small proportion of all possible damage in a sewer. The inspection outcome of such systems relies not only on the quality of the acquired images, but also on the off-line. recognition and classification conducted by human operators. In consequence, CCTV-based platforms are frequently not effective. Infrared, microwave, optical, and ultrasonic-based sensors have been proposed to complement the existing CCTV-based approach and to improve inspection results. New inspection devices employing multiple sensors and being capable of carrying out remote sewer inspection tasks are under research.

Pipe inspection using a laser-based transducer and automated analysis techniques

This paper presents a new sensing methodology for the automated inspection of pipes. Standard inspection systems, as they are for example used in waste pipes and drains, are based on closed-circuit television cameras which are mounted on remotely controlled platforms and connected to remote video recording facilities. Two of the main disadvantages of such camera-based inspection systems are: 1) the poor quality of the acquired images due to difficult lighting conditions and 2) the susceptibility to error during the offline video assessment conducted by human operators. The objective of this research is to overcome these disadvantages and to create an intelligent sensing approach for improved and automated pipe-condition assessment. This approach makes use of a low-cost lighting profiler and a camera which acquires images of the light projections on the pipe wall. A novel method for extracting and analyzing intensity variations in the acquired images is introduced. The image data analysis is based on differential processing leading to highly-noise tolerant algorithms, particularly well suited for the detection of small faults in harsh environments. With the subsequent application of artificial neural networks, the system is capable of recognizing defective areas with a high success rate. Experiments in a range of waste pipes with different diameters and material properties have been conducted and test results are presented.

Automated Pipe Defect Detection and Categorization Using Camera/Laser-Based Profiler and Artificial Neural Network

Closed-circuit television (CCTV) is currently used in many inspection applications, such as the inspection of nonaccessible pipe surfaces. This human-oriented approach based on offline analysis of the raw images is highly subjective and prone to error because of the exorbitant amount of data to be assessed. Laser profilers have been recently proposed to project well-defined light patterns, improving the illumination of standard CCTV systems as well as enhancing the capability of automating the assessment process. This research shows that positional (geometrical) as well as intensity information, related to potential defects, can be extracted from the acquired laser projections. While most researchers focus on the analysis of positional information obtained from the acquired profiler signals, here the intensity information contained within the reflected light is also exploited for the purpose of defect classification and visualization. This paper describes novel strategies created for the automation of defect classification in tubular structures and explores new methods to fuse intensity and positional information, achieving improved multivariable defect classification. The acquired camera/laser images are processed in order to extract signal information for the purpose of visualization and map creation for further assessment. Then, a two-stage approach based on image processing and artificial neural networks is used to classify the images. First, a binary classifier identifies defective pipe sections, and then in a second stage, the defects are classified into different types, such as holes, cracks, and protruding obstacles. Experimental results are provided. Note to Practitioners-The method presented in this paper aims to automate the inspection of nonaccessible pipe surfaces. The method was thought to be employed in the inspection of sewers; however, it could be used in many other industrial applications and could also be extended to other shapes rather than tubular structures. A laser ring profiler, consisting, for instance, of a laser diode and a ring projector, can be easily integrated into existing closed-circuit television systems. The proposed algorithm identifies defective areas and categorizes the types of defects, analyzing the successive recorded camera images that will contain the reflected ring of light. The algorithm, that can be used online, makes use of the deformation of the reflected laser ring together with its changes in intensity. The fact of combining the two kinds of data using artificial-intelligent algorithms makes the method robust enough to work in harsh environments

A Time-Efficient Method for Anomaly Detection in Hyperspectral Images

We propose a computationally efficient method for determining anomalies in hyperspectral data. In the first stage of the algorithm, the background classes, which are the dominant classes in the image, are found. The method consists of robust clustering of a randomly chosen small percentage of the image pixels. The clusters are the representatives of the background classes. By using a subset of the pixels instead of the whole image, the computation is sped up, and the probability of including outliers in the background model is reduced. Anomalous pixels are the pixels with spectra that have large relative distances from the cluster centers. Several clustering techniques are investigated, and experimental results using realistic hyperspectral data are presented. A self-organizing map clustered using the local minima of the U-matrix (unified distance matrix) is identified as the most reliable method for background class extraction. The proposed algorithm for anomaly detection is evaluated using realistic hyperspectral data, is compared with a state-of-the-art anomaly detection algorithm, and is shown to perform significantly better.

Automated sewer pipe inspection through image processing

An innovative inspection method to assess the condition of sewer pipes is proposed in this paper. The standard sewer inspection technique, based on closed-circuit television systems, has a relatively poor performance; a video camera is mounted on a robot and the video recording is provided off-line to an engineer who classifies any defects. The focus of this research is the automated identification and location of discontinuities in the internal surface of sewers. The transducer used is an assembly of a CCD camera and optical elements to generate a ring-shaped laser pattern. The automated inspection method consists of several stages including the segmentation of the image into characteristic geometric features and potential defect regions. Automatic recognition, rating and classification of pipe defects are carried out by means of the computation of a partial histogram based on adaptive image processing techniques. Experiments in a realistic environment have been conducted and results are presented.

Spectral Unmixing With Negative and Superunity Abundances for Subpixel Anomaly Detection

We propose a low false alarm methodology to determine anomalies in hyperspectral data. The method is based on the assumptions that the linear mixing model is valid and that, due to the resolution of the image, most pixels are mixtures of common substances, of which pure pixels (not mixtures) are rare. In the first stage of the algorithm, the classes associated with the background, which are the dominant classes in the image, are found by clustering the image pixels. The resulting clusters may be considered as representatives of the background classes in the image. In order to determine the anomalous pixels, a threshold may be applied to the distance between the pixel spectrum and the cluster centers. However, pixels corresponding to anomalies and pure substances will both show high distances. If we consider that the background classes are themselves most likely mixtures of other materials, the pixels within the convex hull formed by the background classes will have positive fractions that are smaller than one. The pure substances, however, will be outside such a convex hull and will show negative or superunity fractions. Pixels with such mixing proportions are explained as linear combinations of the background classes and, therefore, as not true anomalies. Pixels corresponding to anomalies, however, when expressed as linear combinations of the background classes, show high residual error even with negative and superunity mixing proportions. We use the unmixing spectral linear model without the nonnegativity constraint to distinguish between false anomalies corresponding to pure substances and real man-made anomalies.

Robust Endmember Extraction in the Presence of Anomalies

Endmember extraction is usually based on the solution of a system of linear equations that allows the identification of some basic spectra in terms of which the observed mixed spectra may be expressed as linear combinations. In this paper, we propose to close the loop of such an approach by identifying only the basic spectra that reproduce the dominant cover classes of a region as endmembers, and distinguishing them from outlier spectra present in the scene. The latter are often confused by other methods as endmember spectra, whereas in many practical applications, they are treated as anomalies or targets in the scene. Thus, the proposed method identifies endmembers in a robust way, separating them from outliers.

A time-efficient clustering method for pure class selection

In order to detect a target or anomaly in a hyper-spectral image the classes associated with the background have to be identified. We propose a computationally efficient methodology to determine the background classes present in the image. The method is based on the assumption that mixed and anomaly pixels are relatively rare in comparison with the abundance of the background class pixels. The method considers the background classes as groups of distinct measurements and consists of robust clustering of a randomly picked small percentage of the image pixels. The resulting clusters may be considered as representatives of the background of the image. Several clustering techniques are investigated and experimental results using hyperspectral data are presented. The proposed technique using a self-organising map is then compared with a state-of the art endmember extraction technique.

Experiments using a laser-based transducer and automated analysis techniques for pipe inspection

This paper presents the experimental results of an automated sensor system for the inspection of tubular structures. The method is applied to the autonomous inspection of sewers overcoming the drawbacks of standard CCTV-based inspection systems. The transducer consists of a low-cost laser-based profiler attached to a standard CCTV camera. Image analysis techniques and artificial neural networks are used to automatically locate and classify the defects in the pipe using the intensity distribution in the acquired camera images. A wide range of tests using data from different types of pipes in realistic conditions have been conducted and are presented here. It is shown that the proposed inspection approach is particularly well suited to complement existing CCTV inspection systems, providing automated and reliable detection of pipe defects in the millimeter range.

Subpixel temporal spectral imaging

Abstract Hyperspectral and multispectral sensors are becoming more accessible. However, their use is still limited to certain civil and military applications such as satellite surveillance. Real-time methodologies for predicting target tracking combined with spectral data has many potential applications, including surveillance, fast moving object tracking, etc. However, tracking small objects in dense scenes is difficult with the assumption of coarse pixels and often requires the use of subpixel detection algorithms, and in particular the use of the linear mixture model. Here we propose to introduce the dimension of time to the problem of spectral unmixing to identify and track sub-pixel size targets. Results are provided within a particle filter framework for simulated and real scenes and compared with previously proposed methods for endmember extraction and tracking in spectral data.