Asynchronous Multi-Object Tracking with an Event Camera

Event cameras are bioinspired vision sensors that output pixel-level brightness changes instead of standard intensity frames. They offer significant advantages over standard cameras, namely a very high dynamic range, no motion blur, and a latency in the order of microseconds. However, due to the fundamentally different structure of the sensor's output, new algorithms that exploit the high temporal resolution and the asynchronous nature of the sensor are required. Recent work has shown that a continuous-time representation of the event camera pose can deal with the high temporal resolution and asynchronous nature of this sensor in a principled way. In this paper, we leverage such a continuous-time representation to perform visual-inertial odometry with an event camera. This representation allows direct integration of the asynchronous events with microsecond accuracy and the inertial measurements at high frequency. The event camera trajectory is approximated by a smooth curve in the space of rigid-body motions using cubic splines. This formulation significantly reduces the number of variables in trajectory estimation problems. We evaluate our method on real data from several scenes and compare the results against ground truth from a motion-capture system. We show that our method provides improved accuracy over the result of a state-of-the-art visual odometry method for event cameras. We also show that both the map orientation and scale can be recovered accurately by fusing events and inertial data. To the best of our knowledge, this is the first work on visual-inertial fusion with event cameras using a continuous-time framework.

Recently, the event camera has become a popular and promising vision sensor in the research of simultaneous localization and mapping and computer vision owing to its advantages: low latency, high dynamic range, and high temporal resolution. As a basic part of the feature-based SLAM system, the feature tracking method using event cameras is still an open question. In this article, we present a novel asynchronous event feature generation and tracking algorithm operating directly on event-streams to fully utilize the natural asynchronism of event cameras. The proposed algorithm consists of an event-corner detection unit, a descriptor construction unit, and an event feature tracking unit. The event-corner detection unit addresses a fast and asynchronous corner detector to extract event-corners from event-streams. For the descriptor construction unit, we propose a novel asynchronous gradient descriptor inspired by the scale-invariant feature transform descriptor, which helps to achieve quantitative measurement of similarity between event feature pairs. The construction of the gradient descriptor can be decomposed into three stages: speed-invariant time surface maintenance and extraction, principal orientation calculation, and descriptor generation. The event feature tracking unit combines the constructed gradient descriptor and an event feature matching method to achieve asynchronous feature tracking. We implement the proposed algorithm in C++ and evaluate it on a public event dataset. The experimental results show that our proposed method achieves improvement in terms of tracking accuracy and real-time performance when compared with the state-of-the-art asynchronous event-corner tracker and with no compromise on the feature tracking lifetime.

Event cameras are bio-inspired sensors that differ from conventional frame cameras: Instead of capturing images at a fixed rate, they asynchronously measure per-pixel brightness changes, and output a stream of events that encode the time, location and sign of the brightness changes. Event cameras offer attractive properties compared to traditional cameras: high temporal resolution (in the order of is), very high dynamic range (140dB vs. 60dB), low power consumption, and high pixel bandwidth (on the order of kHz) resulting in reduced motion blur. Hence, event cameras have a large potential for robotics and computer vision in challenging scenarios for traditional cameras, such as low-latency, high speed, and high dynamic range. However, novel methods are required to process the unconventional output of these sensors in order to unlock their potential. This paper provides a comprehensive overview of the emerging field of event-based vision, with a focus on the applications and the algorithms developed to unlock the outstanding properties of event cameras. We present event cameras from their working principle, the actual sensors that are available and the tasks that they have been used for, from low-level vision (feature detection and tracking, optic flow, etc.) to high-level vision (reconstruction, segmentation, recognition). We also discuss the techniques developed to process events, including learning-based techniques, as well as specialized processors for these novel sensors, such as spiking neural networks. Additionally, we highlight the challenges that remain to be tackled and the opportunities that lie ahead in the search for a more efficient, bio-inspired way for machines to perceive and interact with the world.

The framed-based camera is commonly used for object tracking tasks. However, under poor light conditions, the frame camera cannot work properly and provide stable detection for the tracking system. Also, it has motion blur when the object is moving fast. Compared to the frame-based camera, the event camera has a higher temporal resolution and dynamic range. The event data is transmitted immediately when the brightness of each pixel changes. The higher temporal resolution helps the event camera avoid the motion blur and provide more precise temporal information. Moreover, the higher dynamic range allows the event camera to work in extreme light conditions, such as evening, raining, and strong sunshine environments. In all, those advantages make the event cameras an applicable complementary sensor to the frame cameras. In this paper, several optical flow algorithms for the event data are tested. Then, a clustering algorithm is proposed to generate the detection and hybrid it with the detection from frame images. At last, the PMBM filter is used to realize object tracking and test it on our self-recorded dataset with the DAVIS346 and several publicly available datasets.KeywordsSensor fusionEvent-based cameraOptical flowObject tracking

The multi-object tracking (MOT) algorithms of the joint detection and embedding (JDE) approach estimate bounding boxes and re-identification (re-ID) features of objects with the single network, which balance the tracking accuracy and inference speed. However, when the appearance information between different objects is highly similar, these algorithms are usually easy to cause identity switches, and the comprehensive tracking performance is poor in crowded scenes. Aiming at the above problems, we propose a stronger multi-object tracking algorithm termed as ReIMOT, based on FairMOT. A joint loss function of combining normalized Softmax Loss and the center distance penalty term is designed to supervise the re-ID branch, which increases the intra-class similarity and makes the extracted appearance features more discriminative. To further improve the tracking performance, we introduce coordinate attention to make the encoder-decoder network focus more on features of interest. The experimental results show that the proposed ReIMOT is more effective than the other advanced multi-object tracking algorithms, and decreases the number of ID switches by 13.8% compared to FairMOT on the MOT17 dataset.

Event cameras with high dynamic range and temporal resolution, which are bio-inspired vision sensors, have shown great potential in event-based tracking tasks, particularly in scenarios involving rapid motion and low levels of illumination. Nonetheless, the efficient extraction of sparse information from event camera remains a persistent challenge. Meanwhile, the event camera works asynchronously, generating a continuous stream of events, rendering it highly compatible with Spiking Neural Networks (SNNs) due to their event-driven nature and low power consumption. Motivated by the issues mentioned above, we propose an Efficient Hybrid Cascade Tracker ( EHCT ) with SNN for object tracking in the event domain. We combine the transformer, convolutional network, and SNN structure skillfully to form the basic Hybrid CNN-SNN-Transformer (HCST) block structure, which is also the central component part of our EHCT network. The HCST block is primarily utilized to process the incoming event data and extract information from both local and global contexts. After several cascade HCST blocks, these two types of information will be efficiently integrated with the preprocessed raw data, which are then fed into the Classifier and Regressor head to produce the bounding box (bbox) of the tracked target. Extensive experiments on various event and RGB frame-based datasets demonstrated that our proposed EHCT algorithm outperforms most of the existing state-of-the-art trackers by a significant margin and also achieves a great advantage in terms of energy consumption. Our source code will be available at https://github.com/masac11/EHCT .

Cameras are appealing sensors for mobile robots because they are small, passive, inexpensive and provide rich information about the environment. While cameras have been used successfully on a plenitude of robots, such as autonomous cars or drones, serious challenges remain: power consumption, latency, dynamic range, and frame rate, among others. The sequences of images acquired by a camera are highly redundant (both in space and time), and both acquiring and processing such an amount of data consumes significant power. This limits the operation time of mobile robots and, moreover, defines a fundamental power-latency tradeoff. Specialized cameras designed for high-speed or high-dynamic-range scenarios are expensive, heavy, and require additional power, which prevents their use in agile mobile robots. In this thesis, we investigate event cameras as a biologically-inspired alternative to overcome the limitations of standard cameras. These neuromorphic vision sensors work in a completely different way: instead of providing a sequence of images (i.e., frames) at a constant rate, event cameras transmit only information from those pixels that undergo a significant brightness change. These pixel-level brightness changes, called events, are timestamped with micro-second resolution and transmitted asynchronously at the time they occur. Hence, event cameras are power efficient because they convey only non-redundant information, and are able to capture very high-speed motions, thus they directly address the power-latency tradeoff. Additionally, event cameras achieve a dynamic range of more than 140dB, compared to about 60dB of standard cameras, because each pixel is autonomous and operates at its own set-point. However, since the output of an event camera is fundamentally different from that of standard cameras for which computer-vision algorithms have been developed during the past fifty years, new algorithms that can deal with the asynchronous nature of the sensor and exploit its high temporal resolution are required to unlock its potential. This thesis presents algorithms for using event cameras in the context of robotics. Since event cameras are novel sensors that are being intensively prototyped and have been commercially available only recently (ca. 2008), the literature on event-based algorithms is scarce. This poses some operational challenges as well as uncountable opportunities to explore in research. This thesis focuses on exploring the possibilities that event cameras bring to some fundamental problems in robotics and computer vision, such as localization and actuation. Amongst others, this thesis provides contributions to solving the localization problem, i.e., for a robot equipped with an event camera to be able to infer its location with respect to a given map of the environment. Classical approaches for robot localization build upon lower-level vision algorithms, and so, this thesis also presents contributions in the topics of detection, extraction, and tracking of salient visual features with an event camera, whose applicability expands far beyond the localization problem. This thesis also presents contributions in the use of event cameras for actuation and closed-loop control, i.e., in endowing the robot with the capabilities to interact with the environment to fulfill a given task. Additionally, this thesis also presents the infrastructure developed to work with event cameras in a de-facto standard robotics platform. The following is a list of contributions: * Software infrastructure, consisting of publicly available drivers, calibration tools, sensor delay characterization, and the first event camera dataset and simulator tailored for 6-DOF (degrees of freedom) camera pose estimation and SLAM (Simultaneous localization and mapping). * We introduce the concept of event and provide an algorithm to compute it. The lifetime endows the events with a finite temporal extent for a proper continuous representation of events in time. * The first method to extract FAST-like visual features (i.e., interest points or corners) from the output of an event camera. The detector operates an order of magnitude faster than previous corner detectors. * The first method to extract and track features from the output of a DAVIS camera (an event camera that also outputs standard frames from the same pixel array). Using these feature tracks, we developed the first sparse, feature-based visual-odometry pipeline. * The first two methods to track the 6-DOF pose of an event camera in a known map.While the first method minimizes the reprojection error of the events and only works on black-and-white scenes consisting of line segments, the second method uses a probabilistic filtering framework that allows tracking at high speeds on natural scenes. * The first application of a continuous-time framework to estimate the trajectory of an event camera, possibly incorporating inertial measurements, showing superior performance over pose-tracking-only methods. * An application of event cameras to collision avoidance of a quadrotor, showing how event cameras can be used to control a robot with very low latency. * An application of the use of an event camera for human-vs-machine slot-car racing, showing that event-driven algorithms are power efficient and can outperform human control.

Event cameras, offering extremely high temporal resolution and high dynamic range, have brought a new perspective to addressing common object detection challenges (e.g., motion blur and low light). However, how to learn a better spatio-temporal representation and exploit rich temporal cues from asynchronous events for object detection still remains an open issue. To address this problem, we propose a novel asynchronous spatio-temporal memory network (ASTMNet) that directly consumes asynchronous events instead of event images prior to processing, which can well detect objects in a continuous manner. Technically, ASTMNet learns an asynchronous attention embedding from the continuous event stream by adopting an adaptive temporal sampling strategy and a temporal attention convolutional module. Besides, a spatio-temporal memory module is designed to exploit rich temporal cues via a lightweight yet efficient inter-weaved recurrent-convolutional architecture. Empirically, it shows that our approach outperforms the state-of-the-art methods using the feed-forward frame-based detectors on three datasets by a large margin (i.e., 7.6% in the KITTI Simulated Dataset, 10.8% in the Gen1 Automotive Dataset, and 10.5% in the 1Mpx Detection Dataset). The results demonstrate that event cameras can perform robust object detection even in cases where conventional cameras fail, e.g., fast motion and challenging light conditions.

Event cameras are novel sensors that report brightness changes in the form of a stream of asynchronous "events" instead of intensity frames. They offer significant advantages with respect to conventional cameras: high temporal resolution, high dynamic range, and no motion blur. While the stream of events encodes in principle the complete visual signal, the reconstruction of an intensity image from a stream of events is an ill-posed problem in practice. Existing reconstruction approaches are based on hand-crafted priors and strong assumptions about the imaging process as well as the statistics of natural images. In this work we propose to learn to reconstruct intensity images from event streams directly from data instead of relying on any hand-crafted priors. We propose a novel recurrent network to reconstruct videos from a stream of events, and train it on a large amount of simulated event data. During training we propose to use a perceptual loss to encourage reconstructions to follow natural image statistics. We further extend our approach to synthesize color images from color event streams. Our quantitative experiments show that our network surpasses state-of-the-art reconstruction methods by a large margin in terms of image quality ( ), while comfortably running in real-time. We show that the network is able to synthesize high framerate videos ( frames per second) of high-speed phenomena (e.g., a bullet hitting an object) and is able to provide high dynamic range reconstructions in challenging lighting conditions. As an additional contribution, we demonstrate the effectiveness of our reconstructions as an intermediate representation for event data. We show that off-the-shelf computer vision algorithms can be applied to our reconstructions for tasks such as object classification and visual-inertial odometry and that this strategy consistently outperforms algorithms that were specifically designed for event data. We release the reconstruction code, a pre-trained model and the datasets to enable further research.

Traditional visual place recognition (VPR), usually using standard cameras, is easy to fail due to glare or high-speed motion. By contrast, event cameras have the advantages of low latency, high temporal resolution, and high dynamic range, which can deal with the above issues. Nevertheless, event cameras are prone to failure in motionless scenes, while standard cameras can still provide appearance information in this case. Thus, exploiting the complementarity of standard cameras and event cameras can effectively improve the performance of VPR algorithms. In the paper, we propose FE-Fusion-VPR, an attention-based multi-scale network architecture for VPR by fusing frames and events. First, the intensity frame and event volume are fed into the two-stream feature extraction network for shallow feature fusion. Next, the three-scale features are obtained through the multi-scale fusion network and aggregated into three sub-descriptors using the VLAD layer. Finally, the weight of each sub-descriptor is learned through the descriptor re-weighting network to obtain the final refined descriptor. Experimental results show that our FE-Fusion-VPR outperforms existing frame-based, event-based and fusion-based VPR methods in most cases on Brisbane-Event-VPR and DDD20 datasets. In a word, compared to the previous works, our FE-Fusion-VPR achieves new state-of-the-art (SOTA) VPR performance on Brisbane-Event-VPR and DDD20 datasets by fusing frames and events.

This paper proposes an asynchronous particle filter to reconstruct high dynamic range (HDR) video from low dynamic range (LDR) frame and event data. Event cameras have the characteristics of HDR and high temporal resolution, but at the same time they discard the absolute intensity information of the image. Traditional CMOS image sensors can provide low-frequency image with intensity information. In order to fuse the two kinds of information with different frequencies, we interpolate the image according to the inter-frame events. The HDR image is obtained by combining the two with an asynchronous particle filter (PF). For the any distribution of event noise parameters, PF is not limited by the event noise distribution and has a better estimation performance. By selecting some challenging scenes from the event camera datasets ACD and IJRR, the experimental results show that our algorithm can achieve higher quality image reconsruction.

Event cameras are neuromorphic vision sensors that are able to capture high dynamic range with low latency in microseconds, without motion blur. Their strength lies in the unique representation of data as asynchronous events, enabling detection of scene structures less invariantly from dynamic luminance changes. However, a single event does not represent spatial information, and events must be integrated to translate into meaningful information. Therefore, state-of-the-art deep learning algorithms have focused on reconstructing the original scene from events. However, as environmental variances are also captured throughout events and restored in reconstructed images, simple reconstruction does not help achieving robust visual place recognition. In this paper, we suggest to use reconstructed event edges denoised for place recognition. While brightness wavers with dynamic environmental variances, edge contours only change with gradient magnitude scale. We utilize the high dynamic range of event cameras to detect these scaled edges from different environments and show that using reconstructed edges shows robust performance in overcoming day-to-night illumination variance without a large training set.

Recent advances in event-based cameras have led to significant developments in robotics, particularly in visual simultaneous localization and mapping (VSLAM) applications. This technique enables real-time camera motion estimation and simultaneous environment mapping using visual sensors on mobile platforms. Event cameras offer several distinct advantages over frame-based cameras, including a high dynamic range, high temporal resolution, low power consumption, and low latency. These attributes make event cameras highly suitable for addressing performance issues in challenging scenarios such as high-speed motion and environments with high-range illumination. This review paper delves into event-based VSLAM (EVSLAM) algorithms, leveraging the advantages inherent in event streams for localization and mapping endeavors. The exposition commences by explaining the operational principles of event cameras, providing insights into the diverse event representations applied in event data preprocessing. A crucial facet of this survey is the systematic categorization of EVSLAM research into three key parts: event preprocessing, event tracking, and sensor fusion algorithms in EVSLAM. Each category undergoes meticulous examination, offering practical insights and guidance for comprehending each approach. Moreover, we thoroughly assess state-of-the-art (SOTA) methods, emphasizing conducting the evaluation on a specific dataset for enhanced comparability. This evaluation sheds light on current challenges and outlines promising avenues for future research, emphasizing the persisting obstacles and potential advancements in this dynamically evolving domain.

Pig husbandry constitutes a significant segment within the broader framework of livestock farming, with porcine well-being emerging as a paramount concern due to its direct implications on pig breeding and production. An easily observable proxy for assessing the health of pigs lies in their daily patterns of movement. The daily movement patterns of pigs can be used as an indicator of their health, in which more active pigs are usually healthier than those who are not active, providing farmers with knowledge of identifying pigs’ health state before they become sick or their condition becomes life-threatening. However, the conventional means of estimating pig mobility largely rely on manual observations by farmers, which is impractical in the context of contemporary centralized and extensive pig farming operations. In response to these challenges, multi-object tracking and pig behavior methods are adopted to monitor pig health and welfare closely. Regrettably, these existing methods frequently fall short of providing precise and quantified measurements of movement distance, thereby yielding a rudimentary metric for assessing pig health. This paper proposes a novel approach that integrates optical flow and a multi-object tracking algorithm to more accurately gauge pig movement based on both qualitative and quantitative analyses of the shortcomings of solely relying on tracking algorithms. The optical flow records accurate movement between two consecutive frames and the multi-object tracking algorithm offers individual tracks for each pig. By combining optical flow and the tracking algorithm, our approach can accurately estimate each pig’s movement. Moreover, the incorporation of optical flow affords the capacity to discern partial movements, such as instances where only the pig’s head is in motion while the remainder of its body remains stationary. The experimental results show that the proposed method has superiority over the method of solely using tracking results, i.e., bounding boxes. The reason is that the movement calculated based on bounding boxes is easily affected by the size fluctuation while the optical flow data can avoid these drawbacks and even provide more fine-grained motion information. The virtues inherent in the proposed method culminate in the provision of more accurate and comprehensive information, thus enhancing the efficacy of decision-making and management processes within the realm of pig farming.

Event cameras are promising sensors that show many advantages over frame-based cameras. Unlike conventional cameras, whose pixels share a common exposure time, event-based cameras represent a novel bio-inspired technology capable of capturing scenes with a high dynamic range and without motion blur. Due to their working principle, an event is generated when a pixel's brightness changes. Therefore, no event data is generated in a scenario where there is no relative motion between the event camera and the scene. However, in this study, we present a new method to enable event generation with a static event camera on a scene with static objects, aiming to eliminate the requirement of relative motion between event camera and the scene. By projecting custom designed grayscale pattern sequences onto static scenes, we successfully triggered a controlled event generation without requiring camera or object motion. Instead of a direct black-to-white transition, we used a sequence of contrast compatible grayscale projection pattern to regulate event rates and prevent bandwidth overload. To prevent event loss over time, we equalized all timestamps to the first timestamp. Since events in event cameras gradually lose their impact and reset over time, this adjustment prevents the decay of event information and ensures a continuous and stable event generation. Despite the absence of motion, we achieved reasonable results in image quality metrics such as MSE, LPIPS, SSIM. In this way, we aim to expand the usage areas of event cameras and make significant progress in data collection processes, especially for static camera and scenes.