Unified detection and tracking of instruments during retinal microsurgery

Methods for tracking an object have generally fallen into two groups: tracking by detection and tracking through local optimization. The advantage of detection-based tracking is its ability to deal with target appearance and disappearance, but it does not naturally take advantage of target motion continuity during detection. The advantage of local optimization is efficiency and accuracy, but it requires additional algorithms to initialize tracking when the target is lost. To bridge these two approaches, we propose a framework for unified detection and tracking as a time-series Bayesian estimation problem. The basis of our approach is to treat both detection and tracking as a sequential entropy minimization problem, where the goal is to determine the parameters describing a target in each frame. To do this we integrate the Active Testing (AT) paradigm with Bayesian filtering, and this results in a framework capable of both detecting and tracking robustly in situations where the target object enters and leaves the field of view regularly. We demonstrate our approach on a retinal tool tracking problem and show through extensive experiments that our method provides an efficient and robust tracking solution.

Range-based Weighted-likelihood Particle Filter for RSS-based Indoor Tracking

Attractive business cases in various application fields contribute to the sustained long-term interest in indoor localization and tracking by the research community. Location tracking is generally treated as a dynamic state estimation problem, consisting of two steps: (i) location estimation through measurement, and (ii) location prediction. For the estimation step, one of the most efficient and low-cost solutions is Received Signal Strength (RSS)-based ranging. However, various challenges - unrealistic propagation model, non-line of sight (NLOS), and multipath propagation - are yet to be addressed. Particle filters are a popular choice for dealing with the inherent non-linearities in both location measurements and motion dynamics. While such filters have been successfully applied to accurate, time-based ranging measurements, dealing with the more error-prone RSS based ranging is still challenging. In this work, we address the above issues with a novel, weighted likelihood, bootstrap particle filter for tracking via RSS-based ranging. Our filter weights the individual likelihoods from different anchor nodes exponentially, according to the ranging estimation. We also employ an improved propagation model for more accurate RSS-based ranging, which we suggested in recent work. We implemented and tested our algorithm in a passive localization system with IEEE 802.15.4 signals, showing that our proposed solution largely outperforms a traditional bootstrap particle filter.

A Baseline Wander Tracking System for Artifact Rejection in Long-Term Electrocardiography

Long-term electrocardiogram (ECG) signals might suffer from relevant baseline disturbances during physical activity. Motion artifacts in particular are more pronounced with dry surface or esophageal electrodes which are dedicated to prolonged ECG recording. In this paper we present a method called baseline wander tracking (BWT) that tracks and rejects strong baseline disturbances and avoids concurrent saturation of the analog front-end. The proposed algorithm shifts the baseline level of the ECG signal to the middle of the dynamic input range. Due to the fast offset shifts, that produce much steeper signal portions than the normal ECG waves, the true ECG signal can be reconstructed offline and filtered using computationally intensive algorithms. Based on Monte Carlo simulations we observed reconstruction errors mainly caused by the non-linearity inaccuracies of the DAC. However, the signal to error ratio of the BWT is higher compared to an analog front-end featuring a dynamic input ranges above 15 mV if a synthetic ECG signal was used. The BWT is additionally able to suppress (electrode) offset potentials without introducing long transients. Due to its structural simplicity, memory efficiency and the DC coupling capability, the BWT is dedicated to high integration required in long-term and low-power ECG recording systems.

Detecting single-target changes in multiple object tracking: The case of peripheral vision.

In sports games, it is often necessary to perceive a large number of moving objects (e.g., the ball and players). In this context, the role of peripheral vision for processing motion information in the periphery is often discussed especially when motor responses are required. In an attempt to test the basal functionality of peripheral vision in those sports-games situations, a Multiple Object Tracking (MOT) task that requires to track a certain number of targets amidst distractors, was chosen. Participants’ primary task was to recall four targets (out of 10 rectangular stimuli) after six seconds of quasi-random motion. As a second task, a button had to be pressed if a target change occurred (Exp 1: stop vs. form change to a diamond for 0.5 s; Exp 2: stop vs. slowdown for 0.5 s). While eccentricities of changes (5-10° vs. 15-20°) were manipulated, decision accuracy (recall and button press correct), motor response time as well as saccadic reaction time were calculated as dependent variables. Results show that participants indeed used peripheral vision to detect changes, because either no or very late saccades to the changed target were executed in correct trials. Moreover, a saccade was more often executed when eccentricities were small. Response accuracies were higher and response times were lower in the stop conditions of both experiments while larger eccentricities led to higher response times in all conditions. Summing up, it could be shown that monitoring targets and detecting changes can be processed by peripheral vision only and that a monitoring strategy on the basis of peripheral vision may be the optimal one as saccades may be afflicted with certain costs. Further research is planned to address the question whether this functionality is also evident in sports tasks.

Detecting single-target changes in multiple object tracking: The case of peripheral vision

In sports games, it is often necessary to perceive a large number of moving objects (e.g., the ball and players). In this context, the role of peripheral vision for processing motion information in the periphery is often discussed especially when motor responses are required. In an attempt to test the capability of using peripheral vision in those sports-games situations, a Multiple-Object-Tracking task that requires to track a certain number of targets amidst distractors, was chosen to determine the sensitivity of detecting target changes with peripheral vision only. Participants’ primary task was to recall four targets (out of 10 rectangular stimuli) after six seconds of quasi-random motion. As a second task, a button had to be pressed if a target change occurred (Exp 1: stop vs. form change to a diamond for 0.5 s; Exp 2: stop vs. slowdown for 0.5 s). Eccentricities of changes (5-10° vs. 15-20°) were manipulated, decision accuracy (recall and button press correct), motor response time and saccadic reaction time (change onset to saccade onset) were calculated and eye-movements were recorded. Results show that participants indeed used peripheral vision to detect changes, because either no or very late saccades to the changed target were executed in correct trials. Moreover, a saccade was more often executed when eccentricities were small. Response accuracies were higher and response times were lower in the stop conditions of both experiments while larger eccentricities led to higher response times in all conditions. Summing up, it could be shown that monitoring targets and detecting changes can be processed by peripheral vision only and that a monitoring strategy on the basis of peripheral vision may be the optimal one as saccades may be afflicted with certain costs. Further research is planned to address the question whether this functionality is also evident in sports tasks.

Impairments of peripheral motion change detection during smooth pursuit eye movements

Introduction: In team sports the ability to use peripheral vision is essential to track a number of players and the ball. By using eye-tracking devices it was found that players either use fixations and saccades to process information on the pitch or use smooth pursuit eye movements (SPEM) to keep track of single objects (Schütz, Braun, & Gegenfurtner, 2011). However, it is assumed that peripheral vision can be used best when the gaze is stable while it is unknown whether motion changes can be equally well detected when SPEM are used especially because contrast sensitivity is reduced during SPEM (Schütz, Delipetkose, Braun, Kerzel, & Gegenfurtner, 2007). Therefore, peripheral motion change detection will be examined by contrasting a fixation condition with a SPEM condition. Methods: 13 participants (7 male, 6 female) were presented with a visual display consisting of 15 white and 1 red square. Participants were instructed to follow the red square with their eyes and press a button as soon as a white square begins to move. White square movements occurred either when the red square was still (fixation condition) or moving in a circular manner with 6 °/s (pursuit condition). The to-be-detected white square movements varied in eccentricity (4 °, 8 °, 16 °) and speed (1 °/s, 2 °/s, 4 °/s) while movement time of white squares was constant at 500 ms. 180 events should be detected in total. A Vicon-integrated eye-tracking system and a button press (1000 Hz) was used to control for eye-movements and measure detection rates and response times. Response times (ms) and missed detections (%) were measured as dependent variables and analysed with a 2 (manipulation) x 3 (eccentricity) x 3 (speed) ANOVA with repeated measures on all factors. Results: Significant response time effects were found for manipulation, F(1,12) = 224.31, p < .01, ηp2 = .95, eccentricity, F(2,24) = 56.43; p < .01, ηp2 = .83, and the interaction between the two factors, F(2,24) = 64.43; p < .01, ηp2 = .84. Response times increased as a function of eccentricity for SPEM only and were overall higher than in the fixation condition. Results further showed missed events effects for manipulation, F(1,12) = 37.14; p < .01, ηp2 = .76, eccentricity, F(2,24) = 44.90; p < .01, ηp2 = .79, the interaction between the two factors, F(2,24) = 39.52; p < .01, ηp2 = .77 and the three-way interaction manipulation x eccentricity x speed, F(2,24) = 3.01; p = .03, ηp2 = .20. While less than 2% of events were missed on average in the fixation condition as well as at 4° and 8° eccentricity in the SPEM condition, missed events increased for SPEM at 16 ° eccentricity with significantly more missed events in the 4 °/s speed condition (1 °/s: M = 34.69, SD = 20.52; 2 °/s: M = 33.34, SD = 19.40; 4 °/s: M = 39.67, SD = 19.40). Discussion: It could be shown that using SPEM impairs the ability to detect peripheral motion changes at the far periphery and that fixations not only help to detect these motion changes but also to respond faster. Due to high temporal constraints especially in team sports like soccer or basketball, fast reaction are necessary for successful anticipation and decision making. Thus, it is advised to anchor gaze at a specific location if peripheral changes (e.g. movements of other players) that require a motor response have to be detected. In contrast, SPEM should only be used if a single object, like the ball in cricket or baseball, is necessary for a successful motor response. References: Schütz, A. C., Braun, D. I., & Gegenfurtner, K. R. (2011). Eye movements and perception: A selective review. Journal of Vision, 11, 1-30. Schütz, A. C., Delipetkose, E., Braun, D. I., Kerzel, D., & Gegenfurtner, K. R. (2007). Temporal contrast sensitivity during smooth pursuit eye movements. Journal of Vision, 7, 1-15.

Detecting single-target changes in multiple object tracking: The case of peripheral vision

In the current study it is investigated whether peripheral vision can be used to monitor multi-ple moving objects and to detect single-target changes. For this purpose, in Experiment 1, a modified MOT setup with a large projection and a constant-position centroid phase had to be checked first. Classical findings regarding the use of a virtual centroid to track multiple ob-jects and the dependency of tracking accuracy on target speed could be successfully replicat-ed. Thereafter, the main experimental variations regarding the manipulation of to-be-detected target changes could be introduced in Experiment 2. In addition to a button press used for the detection task, gaze behavior was assessed using an integrated eye-tracking system. The anal-ysis of saccadic reaction times in relation to the motor response shows that peripheral vision is naturally used to detect motion and form changes in MOT because the saccade to the target occurred after target-change offset. Furthermore, for changes of comparable task difficulties, motion changes are detected better by peripheral vision than form changes. Findings indicate that capabilities of the visual system (e.g., visual acuity) affect change detection rates and that covert-attention processes may be affected by vision-related aspects like spatial uncertainty. Moreover, it is argued that a centroid-MOT strategy might reduce the amount of saccade-related costs and that eye-tracking seems to be generally valuable to test predictions derived from theories on MOT. Finally, implications for testing covert attention in applied settings are proposed.

Video analysis based vehicle detection and tracking using an MCMC sampling framework

This article presents a probabilistic method for vehicle detection and tracking through the analysis of monocular images obtained from a vehicle-mounted camera. The method is designed to address the main shortcomings of traditional particle filtering approaches, namely Bayesian methods based on importance sampling, for use in traffic environments. These methods do not scale well when the dimensionality of the feature space grows, which creates significant limitations when tracking multiple objects. Alternatively, the proposed method is based on a Markov chain Monte Carlo (MCMC) approach, which allows efficient sampling of the feature space. The method involves important contributions in both the motion and the observation models of the tracker. Indeed, as opposed to particle filter-based tracking methods in the literature, which typically resort to observation models based on appearance or template matching, in this study a likelihood model that combines appearance analysis with information from motion parallax is introduced. Regarding the motion model, a new interaction treatment is defined based on Markov random fields (MRF) that allows for the handling of possible inter-dependencies in vehicle trajectories. As for vehicle detection, the method relies on a supervised classification stage using support vector machines (SVM). The contribution in this field is twofold. First, a new descriptor based on the analysis of gradient orientations in concentric rectangles is dened. This descriptor involves a much smaller feature space compared to traditional descriptors, which are too costly for real-time applications. Second, a new vehicle image database is generated to train the SVM and made public. The proposed vehicle detection and tracking method is proven to outperform existing methods and to successfully handle challenging situations in the test sequences.

Efficient Model-based 3D Tracking by Using Direct Image Registration

This thesis deals with the problem of efficiently tracking 3D objects in sequences of images. We tackle the efficient 3D tracking problem by using direct image registration. This problem is posed as an iterative optimization procedure that minimizes a brightness error norm. We review the most popular iterative methods for image registration in the literature, turning our attention to those algorithms that use efficient optimization techniques. Two forms of efficient registration algorithms are investigated. The first type comprises the additive registration algorithms: these algorithms incrementally compute the motion parameters by linearly approximating the brightness error function. We centre our attention on Hager and Belhumeur’s factorization-based algorithm for image registration. We propose a fundamental requirement that factorization-based algorithms must satisfy to guarantee good convergence, and introduce a systematic procedure that automatically computes the factorization. Finally, we also bring out two warp functions to register rigid and nonrigid 3D targets that satisfy the requirement. The second type comprises the compositional registration algorithms, where the brightness function error is written by using function composition. We study the current approaches to compositional image alignment, and we emphasize the importance of the Inverse Compositional method, which is known to be the most efficient image registration algorithm. We introduce a new algorithm, the Efficient Forward Compositional image registration: this algorithm avoids the necessity of inverting the warping function, and provides a new interpretation of the working mechanisms of the inverse compositional alignment. By using this information, we propose two fundamental requirements that guarantee the convergence of compositional image registration methods. Finally, we support our claims by using extensive experimental testing with synthetic and real-world data. We propose a distinction between image registration and tracking when using efficient algorithms. We show that, depending whether the fundamental requirements are hold, some efficient algorithms are eligible for image registration but not for tracking.

Vision-based vehicle detection and tracking with a mobile camera using a statistical framework

En esta tesis se aborda la detección y el seguimiento automático de vehículos mediante técnicas de visión artificial con una cámara monocular embarcada. Este problema ha suscitado un gran interés por parte de la industria automovilística y de la comunidad científica ya que supone el primer paso en aras de la ayuda a la conducción, la prevención de accidentes y, en última instancia, la conducción automática. A pesar de que se le ha dedicado mucho esfuerzo en los últimos años, de momento no se ha encontrado ninguna solución completamente satisfactoria y por lo tanto continúa siendo un tema de investigación abierto. Los principales problemas que plantean la detección y seguimiento mediante visión artificial son la gran variabilidad entre vehículos, un fondo que cambia dinámicamente debido al movimiento de la cámara, y la necesidad de operar en tiempo real. En este contexto, esta tesis propone un marco unificado para la detección y seguimiento de vehículos que afronta los problemas descritos mediante un enfoque estadístico. El marco se compone de tres grandes bloques, i.e., generación de hipótesis, verificación de hipótesis, y seguimiento de vehículos, que se llevan a cabo de manera secuencial. No obstante, se potencia el intercambio de información entre los diferentes bloques con objeto de obtener el máximo grado posible de adaptación a cambios en el entorno y de reducir el coste computacional. Para abordar la primera tarea de generación de hipótesis, se proponen dos métodos complementarios basados respectivamente en el análisis de la apariencia y la geometría de la escena. Para ello resulta especialmente interesante el uso de un dominio transformado en el que se elimina la perspectiva de la imagen original, puesto que este dominio permite una búsqueda rápida dentro de la imagen y por tanto una generación eficiente de hipótesis de localización de los vehículos. Los candidatos finales se obtienen por medio de un marco colaborativo entre el dominio original y el dominio transformado. Para la verificación de hipótesis se adopta un método de aprendizaje supervisado. Así, se evalúan algunos de los métodos de extracción de características más populares y se proponen nuevos descriptores con arreglo al conocimiento de la apariencia de los vehículos. Para evaluar la efectividad en la tarea de clasificación de estos descriptores, y dado que no existen bases de datos públicas que se adapten al problema descrito, se ha generado una nueva base de datos sobre la que se han realizado pruebas masivas. Finalmente, se presenta una metodología para la fusión de los diferentes clasificadores y se plantea una discusión sobre las combinaciones que ofrecen los mejores resultados. El núcleo del marco propuesto está constituido por un método Bayesiano de seguimiento basado en filtros de partículas. Se plantean contribuciones en los tres elementos fundamentales de estos filtros: el algoritmo de inferencia, el modelo dinámico y el modelo de observación. En concreto, se propone el uso de un método de muestreo basado en MCMC que evita el elevado coste computacional de los filtros de partículas tradicionales y por consiguiente permite que el modelado conjunto de múltiples vehículos sea computacionalmente viable. Por otra parte, el dominio transformado mencionado anteriormente permite la definición de un modelo dinámico de velocidad constante ya que se preserva el movimiento suave de los vehículos en autopistas. Por último, se propone un modelo de observación que integra diferentes características. En particular, además de la apariencia de los vehículos, el modelo tiene en cuenta también toda la información recibida de los bloques de procesamiento previos. El método propuesto se ejecuta en tiempo real en un ordenador de propósito general y da unos resultados sobresalientes en comparación con los métodos tradicionales. ABSTRACT This thesis addresses on-road vehicle detection and tracking with a monocular vision system. This problem has attracted the attention of the automotive industry and the research community as it is the first step for driver assistance and collision avoidance systems and for eventual autonomous driving. Although many effort has been devoted to address it in recent years, no satisfactory solution has yet been devised and thus it is an active research issue. The main challenges for vision-based vehicle detection and tracking are the high variability among vehicles, the dynamically changing background due to camera motion and the real-time processing requirement. In this thesis, a unified approach using statistical methods is presented for vehicle detection and tracking that tackles these issues. The approach is divided into three primary tasks, i.e., vehicle hypothesis generation, hypothesis verification, and vehicle tracking, which are performed sequentially. Nevertheless, the exchange of information between processing blocks is fostered so that the maximum degree of adaptation to changes in the environment can be achieved and the computational cost is alleviated. Two complementary strategies are proposed to address the first task, i.e., hypothesis generation, based respectively on appearance and geometry analysis. To this end, the use of a rectified domain in which the perspective is removed from the original image is especially interesting, as it allows for fast image scanning and coarse hypothesis generation. The final vehicle candidates are produced using a collaborative framework between the original and the rectified domains. A supervised classification strategy is adopted for the verification of the hypothesized vehicle locations. In particular, state-of-the-art methods for feature extraction are evaluated and new descriptors are proposed by exploiting the knowledge on vehicle appearance. Due to the lack of appropriate public databases, a new database is generated and the classification performance of the descriptors is extensively tested on it. Finally, a methodology for the fusion of the different classifiers is presented and the best combinations are discussed. The core of the proposed approach is a Bayesian tracking framework using particle filters. Contributions are made on its three key elements: the inference algorithm, the dynamic model and the observation model. In particular, the use of a Markov chain Monte Carlo method is proposed for sampling, which circumvents the exponential complexity increase of traditional particle filters thus making joint multiple vehicle tracking affordable. On the other hand, the aforementioned rectified domain allows for the definition of a constant-velocity dynamic model since it preserves the smooth motion of vehicles in highways. Finally, a multiple-cue observation model is proposed that not only accounts for vehicle appearance but also integrates the available information from the analysis in the previous blocks. The proposed approach is proven to run near real-time in a general purpose PC and to deliver outstanding results compared to traditional methods.

Detection and Tracking of Traffic Signs Using a Recursive Bayesian Decision Framework

In this paper we propose a new method for the automatic detection and tracking of road traffic signs using an on-board single camera. This method aims to increase the reliability of the detections such that it can boost the performance of any traffic sign recognition scheme. The proposed approach exploits a combination of different features, such as color, appearance, and tracking information. This information is introduced into a recursive Bayesian decision framework, in which prior probabilities are dynamically adapted to tracking results. This decision scheme obtains a number of candidate regions in the image, according to their HS (Hue-Saturation). Finally, a Kalman filter with an adaptive noise tuning provides the required time and spatial coherence to the estimates. Results have shown that the proposed method achieves high detection rates in challenging scenarios, including illumination changes, rapid motion and significant perspective distortion

Simultaneous 3D object tracking and camera parameter estimation by Bayesian methods and transdimensional MCMC sampling

Multi-camera 3D tracking systems with overlapping cameras represent a powerful mean for scene analysis, as they potentially allow greater robustness than monocular systems and provide useful 3D information about object location and movement. However, their performance relies on accurately calibrated camera networks, which is not a realistic assumption in real surveillance environments. Here, we introduce a multi-camera system for tracking the 3D position of a varying number of objects and simultaneously refin-ing the calibration of the network of overlapping cameras. Therefore, we introduce a Bayesian framework that combines Particle Filtering for tracking with recursive Bayesian estimation methods by means of adapted transdimensional MCMC sampling. Addi-tionally, the system has been designed to work on simple motion detection masks, making it suitable for camera networks with low transmission capabilities. Tests show that our approach allows a successful performance even when starting from clearly inaccurate camera calibrations, which would ruin conventional approaches.

EVA: Laparoscopic instrument tracking based on endoscopic video analysis for psychomotor skills assessment

INTRODUCTION: The EVA (Endoscopic Video Analysis) tracking system a new tracking system for extracting motions of laparoscopic instruments based on non-obtrusive video tracking was developed. The feasibility of using EVA in laparoscopic settings has been tested in a box trainer setup. METHODS: EVA makes use of an algorithm that employs information of the laparoscopic instrument's shaft edges in the image, the instrument's insertion point, and the camera's optical centre to track the 3D position of the instrument tip. A validation study of EVA comprised a comparison of the measurements achieved with EVA and the TrEndo tracking system. To this end, 42 participants (16 novices, 22 residents, and 4 experts) were asked to perform a peg transfer task in a box trainer. Ten motion-based metrics were used to assess their performance. RESULTS: Construct validation of the EVA has been obtained for seven motion-based metrics. Concurrent validation revealed that there is a strong correlation between the results obtained by EVA and the TrEndo for metrics such as path length (p=0,97), average speed (p=0,94) or economy of volume (p=0,85), proving the viability of EVA. CONCLUSIONS: EVA has been successfully used in the training setup showing potential of endoscopic video analysis to assess laparoscopic psychomotor skills. The results encourage further implementation of video tracking in training setups and in image guided surgery.

Relevance of motion-related assessment metrics in laparoscopic surgery

INTRODUCTION: Motion metrics have become an important source of information when addressing the assessment of surgical expertise. However, their direct relationship with the different surgical skills has not been fully explored. The purpose of this study is to investigate the relevance of motion-related metrics in the evaluation processes of basic psychomotor laparoscopic skills, as well as their correlation with the different abilities sought to measure. METHODS: A framework for task definition and metric analysis is proposed. An explorative survey was first conducted with a board of experts to identify metrics to assess basic psychomotor skills. Based on the output of that survey, three novel tasks for surgical assessment were designed. Face and construct validation study was performed, with focus on motion-related metrics. Tasks were performed by 42 participants (16 novices, 22 residents and 4 experts). Movements of the laparoscopic instruments were registered with the TrEndo tracking system and analyzed. RESULTS: Time, path length and depth showed construct validity for all three tasks. Motion smoothness and idle time also showed validity for tasks involving bi-manual coordination and tasks requiring a more tactical approach respectively. Additionally, motion smoothness and average speed showed a high internal consistency, proving them to be the most task-independent of all the metrics analyzed. CONCLUSION: Motion metrics are complementary and valid for assessing basic psychomotor skills, and their relevance depends on the skill being evaluated. A larger clinical implementation, combined with quality performance information, will give more insight on the relevance of the results shown in this study.

Handmade Task Tracking Applied to Cognitive Rehabilitation

This article presents research focused on tracking manual tasks that are applied in cognitive rehabilitation so as to analyze the movements of patients who suffer from Apraxia and Action Disorganization Syndrome (AADS). This kind of patients find executing Activities of Daily Living (ADL) too difficult due to the loss of memory and capacity to carry out sequential tasks or the impossibility of associating different objects with their functions. This contribution is developed from the work of Universidad Politécnica de Madrid and Technical University of Munich in collaboration with The University of Birmingham. The KinectTM for Windows© device is used for this purpose. The data collected is compared to an ultrasonic motion capture system. The results indicate a moderate to strong correlation between signals. They also verify that KinectTM is very suitable and inexpensive. Moreover, it turns out to be a motion-capture system quite easy to implement for kinematics analysis in ADL.