Detect Target Objects Research Articles

Augmenting Training Samples with a Large Number of Rough Segmentation Datasets

We revisit the problem with generic object recognition from the point of view of human-computer interaction. While many existing algorithms for generic object recognition first try to detect target objects before features are extracted and classified in processing, our work is motivated by the belief that solving the task of detection by computer is not always necessary in many practical situations, such as those involving mobile recognition systems with touch displays and cameras. It is natural for these systems to ask users to input the segmentation data for targets through their touch displays. Speaking from the perspective of usability, such systems should involve rough segmentation to reduce the user workload. In this situation, different people would provide different segmentation data. Here, an interesting question arises - if multiple training samples are generated from a single image by using various segmentation data created by different people, what would happen to the accuracy of classification? We created “20 wild bird datasets” that had a large number of rough segmentation datasets made by 383 people in an attempt to answer this question. Our experiments revealed two interesting facts: (i) generating multiple training samples from a single image had positive effects on classification accuracies, especially when image features including spatial information were used and (ii) augmenting training samples with artificial segmentation data synthesized with a morphing technique also had slightly positive effects on classification accuracies.

Segmentation techniques for target recognition

This paper presents an overview of the methodologies and algorithms for segmenting 2D images as a means in detecting target objects embedded in visual images for an Automatic Target Detection application.

Development of a detector model for generation of synthetic radiographs of cargo containers

Creation of synthetic cargo-container radiographs that possess attributes of their empirical counterparts requires accurate models of the imaging-system response. Synthetic radiographs serve as surrogate data in studies aimed at determining system effectiveness for detecting target objects when it is impractical to collect a large set of empirical radiographs. In the case where a detailed understanding of the detector system is available, an accurate detector model can be derived from first-principles. In the absence of this detail, it is necessary to derive empirical models of the imaging-system response from radiographs of well-characterized objects. Such a case is the topic of this work, where we demonstrate the development of an empirical model of a gamma-ray radiography system with the intent of creating a detector-response model that translates uncollided photon transport calculations into realistic synthetic radiographs. The detector-response model is calibrated to field measurements of well-characterized objects thus incorporating properties such as system sensitivity, spatial resolution, contrast and noise.

Relationship of Vertical Illuminance to Pedestrian Visibility in Crosswalks

Twenty-six participants evaluated a series of crosswalk lighting designs by visually detecting objects at each crosswalk location while traveling in a moving vehicle. The research was performed on a closed test track under nighttime conditions while the participants were driving an SUV with regular halogen headlamps. The conditions included several vertical illuminance levels (6, 10, 20, and 30 lux), varied luminaire types [high-pressure sodium (HPS) and metal halide (MH)], and various target object types (pedestrian and surrogate objects). Only one age group of participants (66 years and older) was used for the study, with equal representation of males and females. The participants were asked to detect objects at each crosswalk location when they were confident an object was present. The results indicated that object detection distances changed on the basis of vertical illuminance level, luminaire type, and object type. Object detection distance for HPS was greatest at 30 vertical lux and for MH at 20 vertical lux. However, these results were moderated by the clothing color of the target object. When object color was considered, pedestrians in white clothing were identified earlier under the HPS lighting condition at 20 lux. Under the MH configuration, denim-clothed objects were detected earlier than black-clothed objects, especially at the 20-lux lighting level. The results suggest that a vertical illuminance level of 20 lux at crosswalk locations provides adequate levels for target object detection. In addition to benefiting from vertical illuminance, target objects that wore white clothing had detection distances superior to other object types of different clothing colors. Recommendations for crosswalk lighting configurations are further discussed.

Ranking factors in peer-to-peer overlay networks

A large number of peer processes are distributed in a peer-to-peer (P2P) overlay network. It is difficult, maybe impossible for a peer to perceive the membership and location of every resource object due to the scalability and openness of a P2P network. In this article, we discuss a fully distributed P2P system where there is no centralized controller. Each peer has to obtain service information from its acquaintance peers and also send its service information to the acquaintance peers. An acquaintance peer of a peer p is a peer about whose service the peer p knows and with which the peer p can directly communicate in an overlay network. Some acquaintance peer might hold obsolete service information and might be faulty. Each peer has to find a more trustworthy one among acquaintance peers. There are many discussions on how to detect peers that hold a target object. However, a peer cannot manipulate an object without being granted access rights (permissions). In addition to detecting what peers hold a target object, we have to find peers granted access rights to manipulate the target object. The trustworthiness of each acquaintance is defined in terms of the satisfiability and ranking factor in this article. The satisfiability of an acquaintance peer shows how much each peer can trust the acquaintance peer through direct communication to not only detect target objects but also obtain their access rights. On the other hand, the ranking factor of an acquaintance peer indicates how much the acquaintance peer is trusted only by trustworthy acquaintance peers which is different from the traditional reputation concept. We evaluate how the trustworthiness of an acquaintance peer is changed through interactions among peers in a detection algorithm.

Abstracts from CyberTherapy 2006

The main characteristic of virtual reality (VR) dwells in carrying spatial information in an analogical mode preserving most of space-time dimensions and interaction modalities that humans usually have in natural contexts. According to this, the spatial nature of virtual environments revealed an advantageous opportunity for the evaluation of residual perceptual abilities and motor-explorative behaviours in patients with cognitive impairment. In particular, it has been suggested that VR can be effectively used to evaluate how cognitive impaired users, such as patients affected by visuo-spatial neglect, are able to explore and memorize the environment in which they are able to interact. Neglect patients present difficulties in attending to stimuli placed in the contralesional space (typically the left hemispace). Several kinds of task can be used to assess and to modulate neglect behavior. These tasks should address the whole range of cognitive domains in which the disorder could be manifested (perception and mental representation, personal/ extrapersonal space, navigation abilities). Since standard evaluation of neglect syndrome is mainly carried out using paper-based tests within the laboratory, the assessment of patient’s behavior impairments in everyday contexts tends to be overlooked. Drawing on these premises, we introduced a Vrsituated approach for the evaluation of neglect patient’s perceptive, memory, and explorative residual abilities in coping with daily contexts. We designed two virtual environments: a small-scale, closed environment within which neglect patients could position objects and navigate; and a large-scale, open one, which patients explored freely. The main goal of the first virtual environment provided is to evaluate patients’ ability in memorizing, recognizing, and replacing objects within all the field of vision. The main aim of the second one is to evaluate patients’ exploration strategies and to analyze their description of context they are engaged with. We tested the environments in four brain-damaged patients. They were requested to immersively explore the environments in order to memorize, replace, and recognize familiar objects. Patients included in the study were previously evaluated with “paper and pencil” neuropsychological assessment which revealed respectively a neglect syndrome, an overcome neglect, a right brain damage without neglect and frontal brain injury (with slight attentional neglect). Results showed that, compared with standard evaluation (that is mainly grounded on the detection and recognition of motionless target objects) patients’ interaction with virtual environments (in which target objects are situated in a dynamical scene) exhibited peculiar explorative behaviors, such as perseverations and/or right-side navigation tendencies, that could be addressed to their specific cognitive impairments. In addition, from a methodological point of view, the introduction of a VR-based assessment allowed us to deliver the complexity of stimulus challenges of naturalistic settings, as well as to monitor them in order to identify what constitute patients’ main troubles in managing daily activities.

Position-, rotation-, scale-, and orientation-invariant multiple object recognition from cluttered scenes

A method of detecting target objects in still images despite any kind of geometrical distortion is demonstrated. Two existing techniques are combined, each one capable of creating invariance to various types of distortion of the target object. A maximum average correlation height (MACH) filter is used to create invariance to orientation and gives good tolerance to background clutter and noise. A log r- mapping is employed to give invariance to in-plane rotation and scale by transforming rotation and scale variations of the target object into vertical and horizontal shifts. The MACH filter is trained on the log r- map of the target for a range of orientations and applied sequentially over regions of interest in the input image. Areas producing a strong correlation response can then be used to determine the position, in-plane rotation, and scale of the target objects in the scene.

Generation of efficient and user-friendly queries for helper robots to detect target objects

We are developing a helper robot that carries out tasks ordered by users through speech. The robot needs a vision system to recognize the objects appearing in the orders. However, conventional vision systems cannot recognize objects in complex scenes. They may find many objects and cannot determine which is the target. This paper proposes a method of using a conversation with the user to solve this problem. The robot asks a question to which the user can easily answer and whose answer can efficiently reduce the number of candidate objects. It considers the characteristics of features used for object identification such as the ease for humans to specify them by word, generating a user-friendly and efficient sequence of questions. Experimental results show that the robot can detect target objects by asking the questions generated by the method.

Searching for “the Top” in Top-Down Control

Although numerous theoretical models implicate the prefrontal cortex (PFC) as a source of top-down control signals, empirical support is limited. In this review, we highlight direct evidence for this view of PFC function and discuss several lines of other supportive findings. Most evidence to-date is only suggestive due to methodological challenges in tracking the spatiotemporal dynamics of cortical networks. However, we discuss several promising approaches that could further our understanding of the role of the PFC in goal-directed behavior.

Why does natural scene categorization require little attention? Exploring attentional requirements for natural and synthetic stimuli

It was recently demonstrated that detecting target objects (e.g., animals) in natural scenes can be done in a dual-task paradigm, in the near absence of spatial attention. Under the same conditions, subjects were unable to perform apparently simpler tasks involving synthetic stimuli (e.g., discriminating randomly rotated Ts and Ls, or a bisected colour disc and its mirror image). Classical theories predict that attention is more critical for the recognition of complex stimuli that cannot be easily separated on a single feature dimension. Therefore, these puzzling results have raised a number of questions. If it is not the complexity of a stimulus, what then determines the recognition task's attentional requirements? How does this differ between natural and artificial stimuli? What can these observations tell us about the mechanism of natural scene processing as well as its relation to attention? Here we show that removing colour information, or doubling the amount of information to be analysed, failed to make the natural scene categorization tasks significantly more “attention demanding”. Conversely, increasing discriminability or predictability did not diminish the need for attention in the case of synthetic stimuli. However, when the familiar letters, such as Ts or Ls, were presented upright, full attention was no longer required for discrimination. This suggests that familiarity and meaningfulness might be among the factors that determine attentional requirements for both natural and synthetic stimuli.

Optical System Design for Thermal Target Recognition by Spiral Scanning [TRSS

Various kinds of systems, that can do target recognition and position detection simultaneously by using infrared sensing detectors, have been developed. In this paper, the detection system TRSS (Thermal target Recognition by Spiral Scanning) adopts linear array shaped uncooled IR detector and uses spiral type fast scanning method for relative position detection of target objects, which radiate an IR region wavelength spectrum. It can detect thermal energy radiating from a 9 m-size target object as far as 200 m distance. And the maximum field of a detector is fully filled with the same size of target object at the minimum approaching distance 50 m. We investigate two types of lens systems. One is a singlet lens and the other is a doublet lens system. Every system includes one aspheric surface and free positioned aperture stop. Many designs of F/1.5 system with <TEX>${\pm}5.2^{\circ}$</TEX> field at the Efl=20, 30 mm conditions for single element and double elements lens system respectively are compared in their resolution performance [MTF] according to the aspheric surface and stop position changing on their optimization process. Optimum design is established including mechanical boundary conditions and manufacturing considerations.

Applying the Multiple Signal Classification Method to Silent Object Detection Using Ambient Noise

The revolutionary concept of using ocean ambient noise positively to detect objects, called acoustic daylight imaging, has attracted much attention. The authors attempted the detection of a silent target object using ambient noise and a wide-band beam former consisting of an array of receivers. In experimental results obtained in air, using the wide-band beam former, we successfully applied the delay-sum array (DSA) method to detect a silent target object in an acoustic noise field generated by a large number of transducers. This paper reports some experimental results obtained by applying the multiple signal classification (MUSIC) method to a wide-band beam former to detect silent targets. The ocean ambient noise was simulated by transducers decentralized to many points in air. Both MUSIC and DSA detected a spherical target object in the noise field. The relative power levels near the target obtained with MUSIC were compared with those obtained by DSA. Then the effectiveness of the MUSIC method was evaluated according to the rate of increase in the maximum and minimum relative power levels.

Passive detection of silent target object by ambient noise: Some experimental results using wideband beamformer by an array of receivers.

Passive detection of silent target object by ambient noise: Some experimental results using wideband beamformer by an array of receivers.