Camouflaged Objects Research Articles

Boundary guidance network for camouflage object detection

Camouflage object detection (COD) aims to detect camouflaged objects hidden in the background region in an image. The difficulty of COD lies in the fact that camouflaged objects are often accompanied with weak boundaries, low contrast, and similar patterns to the background. Although various methods have been proposed to address these challenges, they still suffer from coarse object boundaries. In this work, we design a novel boundary guidance network for COD, which follows a two-step framework: localization and refinement. Firstly, an Initial Localization Decoder is proposed to capture multi-scale cues by embedding a Hierarchical-Split Convolution block. After obtained the coarse localization of the camouflaged object, we further propose a Residual Refinement Decoder to fix the missing object parts and boundary details progressively. Each of the proposed decoder consists of a region branch and a boundary branch for object detection and boundary detection respectively. To sufficiently leverage their complementary features, we design a novel Boundary-Guide-Region module. Benefiting from the guidance of the boundary feature, the region branch can focus on the inside parts of the boundary for residual learning, thus leads to more accurate detection. Extensive experimental results on four benchmark datasets demonstrate that our method outperforms existing state-of-the-art algorithms in both object accuracy and boundary accuracy with real-time speed.

Cascade and Fusion: A Deep Learning Approach for Camouflaged Object Sensing.

The demand for the sensor-based detection of camouflage objects widely exists in biological research, remote sensing, and military applications. However, the performance of traditional object detection algorithms is limited, as they are incapable of extracting informative parts from low signal-to-noise ratio features. To address this problem, we propose Camouflaged Object Detection with Cascade and Feedback Fusion (CODCEF), a deep learning framework based on an RGB optical sensor that leverages a cascaded structure with Feedback Partial Decoders (FPD) instead of a traditional encoder–decoder structure. Through a selective fusion strategy and feedback loop, FPD reduces the loss of information and the interference of noises in the process of feature interweaving. Furthermore, we introduce Pixel Perception Fusion (PPF) loss, which aims to pay more attention to local pixels that might become the edges of an object. Experimental results on an edge device show that CODCEF achieved competitive results compared with 10 state-of-the-art methods.

Inferring Camouflaged Objects by Texture-Aware Interactive Guidance Network

Camouflaged objects, similar to the background, show indefinable boundaries and deceptive textures, which increases the difficulty of detection task and makes the model rely on features with more information. Herein, we design a texture label to facilitate our network for accurate camouflaged object segmentation. Motivated by the complementary relationship between texture labels and camouflaged object labels, we propose an interactive guidance framework named TINet, which focuses on finding the indefinable boundary and the texture difference by progressive interactive guidance. It maximizes the guidance effect of refined multi-level texture cues on segmentation. Specifically, texture perception decoder (TPD) makes a comprehensive analysis of texture information in multiple scales. Feature interaction guidance decoder (FGD) interactively refines multi-level features of camouflaged object detection and texture detection level by level. Holistic perception decoder (HPD) enhances FGD results by multi-level holistic perception. In addition, we propose a boundary weight map to help the loss function pay more attention to the object boundary. Sufficient experiments conducted on COD and SOD datasets demonstrate that the proposed method performs favorably against 23 state-of-the-art methods.

Integrating Part-Object Relationship and Contrast for Camouflaged Object Detection

Object detectors that solely rely on image contrast are struggling to detect camouflaged objects in images because of the high similarity between camouflaged objects and their surroundings. To address this issue, in this paper, we investigate the role of the part-object relationship for camouflaged object detection. Specifically, we propose a Part-Object relationship and Contrast Integrated Network (POCINet) covering both search and identification stages, where each stage adopts an appropriate scheme to engage the contrast information and part-object relational knowledge for camouflaged pattern decoding. Besides, we bridge these two stages via a Search-to-Identification Guidance (SIG) module, in which the search result, as well as decoded semantic knowledge, jointly enhances the features encoding ability of the identification stage. Experimental results demonstrate the superiority of our algorithm on three datasets. Notably, our algorithm raises $F_\beta $ of the best existing method by approximately 17 points on the CPD1K dataset. The source code will be released soon.

A review of modern materials used in military camouflage within the radar frequency range

The article presents an overview of materials which can be employed used to camouflage objects on the modern battlefield in the radar frequency range.

A principled approach to the detection of camouflaged objects

A principled approach to the detection of camouflaged objects

Camouflaged Object Detection and Tracking: A Survey

Moving object detection and tracking have various applications, including surveillance, anomaly detection, vehicle navigation, etc. The literature on object detection and tracking is rich enough, and there exist several essential survey papers. However, the research on camouflage object detection and tracking is limited due to the complexity of the problem. Existing work on this problem has been done based on either biological characteristics of the camouflaged objects or computer vision techniques. In this paper, we review the existing camouflaged object detection and tracking techniques using computer vision algorithms from the theoretical point of view. This paper also addresses several issues of interest as well as future research direction in this area. We hope this paper will help the reader to learn the recent advances in camouflaged object detection and tracking.

Person identification with aerial imaginary using SegNet based semantic segmentation

In recent days, people in remote area suffer a lot due to variety of natural calamities such as flooding, earthquake and so on. It has been noted that people used to stay in top portions of their house when there is a flooding issue. Hence, it is very difficult for the rescue team to identify the location of a person by looking at the parts of a person such as hands, legs and partial image of a face using the existing approaches. In this proposed approach, an idea for detecting person when there is only parts such as legs, hands are visible from remote, wild or non-urban areas with the help of UAV-Unmanned Aerial Vehicle has been suggested. Detecting person and identifying the location from the image tends to be a difficult process due to very small and camouflaged objects in the images collected. In this approach, Semantic Segmentation using deep learning approach has been applied in order to detect a person. SegNet- Segmentation Network is the network architecture used in the process of semantically segment the image according to each pixel, hence identifying person is easy. The main objective of this proposed model is that, sometime UAV image may contain partial person images, like legs, hand, etc., that could not be identified by existing approaches were being recognized and identified successfully. This model is trained and tested using HERIDAL dataset. Over 70% images were trained and 30% images were used for testing. This enhanced deep learning model named as Semantic SegNet model achieved an accuracy of 91.04%. This proposed Semantic SegNet model has been compared with existing approaches such as VGG16, GoogleNet and ResNet- Residual neural Network for the same set of trained and tested images. Comparison table declared that this proposed Semantic SegNet Model outperformed other existing models.

Camouflage detection with texture statistical characterization in autonomous systems

PurposeThe purpose of the research work is to detect camouflaged objects in autonomous systems of military applications and civilian applications such as detecting insects in paddy fields, identifying duplicate products in different texture environments.Design/methodology/approachCamouflaged objects detection is performed by smoothing texture with nonlinear models and characterizing with statistical methods to detect the objects.FindingsThere are few challenges in existing camouflaged objects detection due to the complexities involved in the detection process. This work proposes a constructive approach with texture statistical characterization for camouflage detection. The proposed technique is found to be better than existing methods while assessing its performance using precision and recall.Research limitations/implicationsEven though there is lot of research work carried, there are few challenges for autonomous systems in camouflage detection due to the complexities involved in the detection process such as texture modeling and dynamic background problems and environment conditions for autonomous system.Practical implicationsCamouflage detection finds potential applications in security systems, surveillance, military and autonomous systems. The proposed work is implemented in different environments for camouflage detection.Social implicationsSocial problems such as image acquisition environment, time of day, desert, forest and grass fields of paddy.Originality/valueThe proposed method detects camouflaged objects in autonomous systems where it is applied for images of different kinds. It is found to be effective on images recorded in battlefield and challenging environments.

Unsupervised Moving Object Detection in Complex Scenes Using Adversarial Regularizations

Moving object detection (MOD) is a fundamental step in many high-level vision-based applications, such as human activity analysis, visual object tracking, autonomous vehicles, surveillance, and security. Most of the existing MOD algorithms observe performance degradation in the presence of complex scenes containing camouflage objects, shadows, dynamic backgrounds, and varying illumination conditions, and captured by static cameras. To appropriately handle these challenges, we propose a Generative Adversarial Network (GAN) based on a moving object detection algorithm, called MOD_GAN. In the proposed algorithm, scene-specific GANs are trained in an unsupervised MOD setting, thereby enabling the algorithm to learn generating background sequences using input from uniformly distributed random noise samples. In addition to adversarial loss, during training, norm-based loss in the image space and discriminator feature-space is also minimized between the generated images and the training data. The additional losses enable the generator to learn subtle background details, resulting in a more realistic complex scene generation. During testing, a novel back-propagation based algorithm is used to generate images with statistics similar to the test images. More appropriate random noise samples are searched by directly minimizing the loss function between the test and generated images both in the image and discriminator feature-spaces. The network is not updated in this step; only the input noise samples are iteratively modified to minimize the loss function. Moreover, motion information is used to ensure that this loss is only computed on small-motion pixels. A novel dataset containing outdoor time-lapsed images from dawn to dusk with a full illumination variation cycle is also proposed to better compare the MOD algorithms in outdoor scenes. Accordingly, extensive experiments on five benchmark datasets and comparison with 30 existing methods demonstrate the strength of the proposed algorithm.

Fringe projection decamouflaging

Unveiling the cheating mask of camouflage is of great significance in scientific and military research, but the acquisition of detailed information about camouflaged objects remains extremely difficult. This paper presents a fringe projection decamouflaging (FPDC) approach that can provide the position, outer rim profile, and shadow information about well-camouflaged objects in a complex scenario. FPDC requires no expensive instruments, just one camera and one projector for monitoring. A bionic detection field, namely, a phase-jump field, is established to detect the intrusion of camouflaged objects. Based on local variations in this field, multiple-parameter decamouflaging is achieved for the first time. Simulations and experiments show the superiority of this technique, which has potential applications in counterespionage, discovery of wild animals, and search-and-rescue.

Camouflaged object detection for machine vision applications

Camouflage image means the objects in foreground mingled with background in this context one prime area in machine vision application is identifying camouflaged objects in image for military and civilian needs. This paper proposes camouflage detection to identify one or more target objects in Camouflaged Images. In this work, a constructive approach presented by characterizing entity-texture and statistical modeling of Camouflaged Images in texture smoothing conditions. The Difficulties in this process are Images of different kinds, Atmospheric Turbulence, and Target Size. There are many potential applications where the needs of decamouflaging like: distinguishing original and duplicate in product production, military applications and autonomous systems where detection of objects in Camouflage images is essential. The implementation is carried out in different environments on camouflaged Images. The performance of the proposed technique is compared with existing methods using precision and recall.

Competing for affection: Perceptual fluency and ambiguity solution.

Human perceptual processes are highly efficient and rapidly extract information to enable fast and accurate responses. The fluency of these processes is reinforcing, meaning that easy-to-perceive objects are liked more as a result of misattribution of the reinforcement affect to the object identity. However, some critical processes are disfluent, yet their completion can be reinforcing leading to object preference through a different route. One such example is identification of objects from camouflage. In a series of 5 experiments, we manipulated object contrast and camouflage to explore the relationship between object preference to perceptual fluency and ambiguity solution. We found that perceptual fluency dominated the process of preference assessment when objects are assessed for "liking". That is, easier-to-perceive objects (high-contrast and noncamouflaged) were preferred over harder-to-perceive objects (low-contrast and camouflaged). However, when objects are assessed for "interest", the disfluent yet reinforcing ambiguity solution process overrode the effect of perceptual fluency, resulting in preference for the harder-to-perceive camouflaged objects over the easier-to-perceive noncamouflaged objects. The results have implications for preference and choice in a wide range of contexts by demonstrating the competition between perceptual fluency and ambiguity solution on preference, and by highlighting the critical factor of the form of preference decision. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Object Detection in Camouflaged Environment with Texture Statistical Features

In camouflage image foreground will be hidden in the background image. Camouflage images can be natural and artificial. Detection of such hidden objects becomes difficult for a machine vision system and takes much time to detect and recognize whereas it is not difficult for Human Perception. Detection process involves two phases; feature search that helps in grasping the characteristic entities of an image like colour, shape, texture, pattern etc., and conjunction search which is useful for recognition of clues from multiple features. The background of the camouflage image may be uniform or non-uniform characteristic entities. Different operations can be performed on characteristic entities to make the background disappear in order to detect the foreground image. In this paper, survey on DE camouflaging methods and framework is proposed to detect objects in camouflage environments. Under this framework, textural smoothing followed by statistical characteristics is used to detect the camouflaged objects that will show a better performance based on statistical features. De-camouflaging can be used in war field, where soldiers hide themselves from the enemies with the texture similar to that of their background and decamouflaging is used to reveal the camouflages in the background

Development of a polarimetric infrared space telescope for nanosatellites

The purpose of this work is to develop a low-cost space telescope for Earth remote sensing that can be housed in a 12U CubeSat with a ground sampling distance (GSD) of 60 m at long wavelength infrared. It will use the three-mirror anastigmatic concept and commercial off-the-shelf components whenever possible. The main mission of the instrument is oil spill detection, especially in the ocean environment. Apart from the obvious functionality of this main mission, oil spill detection can be very useful for detecting aircraft that have crashed in the ocean. As secondary missions, this instrument can be used to map deforestation, find lost ships in the ocean, and help fight drug smuggling by finding camouflaged man-made objects. In order to fulfill these missions, the instrument will extract polarization information from the target. The optical design indicates that the optical performance of the instrument has a modulation transfer function >53 % at Nyquist frequency and a native GSD of 60 m and that the instrument is diffraction limited (wavefront error of 0.036 wavelengths and Strehl ratio of 95.77%). The mechanical design indicates that the maximum deflection caused by gravity induces a minimal error (smaller than one over one hundred and forty) when compared to the mean wavelength of operation. The radiometric analyses indicate that the instrument has an NEDT of 56.30 mK.

Anabranch network for camouflaged object segmentation

Camouflaged objects attempt to conceal their texture into the background and discriminating them from the background is hard even for human beings. The main objective of this paper is to explore the camouflaged object segmentation problem, namely, segmenting the camouflaged object(s) for a given image. This problem has not been well studied in spite of a wide range of potential applications including the preservation of wild animals and the discovery of new species, surveillance systems, search-and-rescue missions in the event of natural disasters such as earthquakes, floods or hurricanes. This paper addresses a new challenging problem of camouflaged object segmentation. To address this problem, we provide a new image dataset of camouflaged objects for benchmarking purposes. In addition, we propose a general end-to-end network, called the Anabranch Network, that leverages both classification and segmentation tasks. Different from existing networks for segmentation, our proposed network possesses the second branch for classification to predict the probability of containing camouflaged object(s) in an image, which is then fused into the main branch for segmentation to boost up the segmentation accuracy. Extensive experiments conducted on the newly built dataset demonstrate the effectiveness of our network using various fully convolutional networks.

Detection of People With Camouflage Pattern Via Dense Deconvolution Network

In this letter, we explore the detection of people with camouflage pattern in cluttered natural scenes. First, considering the lack of open evaluation and training data on camouflaged people detection, a specific dataset of camouflaged people in natural scenes is constructed by us for the first time to the best of our knowledge. Secon, due to the serious corruption of the discrimination of low-level features by the camouflage patterns and cluttered background, we extract the high-level semantic features in deep convolution network and introduce short connections in deconvolution phase, to construct the dense deconvolution network. In training procedure, we augment and shift the images of camouflaged people to generate the proper training data. Attributing to the usage and fusion of semantic information, the proposed network effectively labels the camouflaged people regions as a whole. Finally, we use the superpixel segmentation and spatial smoothness constraint for further improvement of the detection result. Experimental results demonstrate that the proposed method outperforms the classical camouflaged object detection method and typical CNN-based detection methods.

Fast satellite imagery of lengthy territories with complex configuration

When solving a number of remote sensing problems, there arose a need for fast satellite imagery of arbitrarily-spaced lengthy territories with complex configuration (borders, roads, rivers, coastlines, etc.), The distinctive features of Earth remote sensing satellites equipped with optical-electronic scanners providing sub-meter spatial resolution are a narrow swath width, which does not allow to take images of arbitrarily-spaced lengthy territories in one-orbit period, and a small pixel size of CCD photocells (i.e., low sensitivity), which does not allow to take images of low-contrast objects, as well as take pictures from neighbouring orbits at low angles of the Sun. An effective technique for controlling the high-resolution remote sensing spacecraft orientation in the process of retargeting (i.e., with non-zero angular velocities) is proposed, which allows to choose the best scan mode and take images of arbitrarily-spaced lengthy territories with complex configuration in one-orbit period, i.e. much faster. Moreover, non-zero angular velocities allow to increase the exposure time and take pictures of low contrast, low light or camouflaged objects. The main characteristics and disadvantages of Earth remote sensing satellites equipped with optical-electronic scanners providing sub-meter spatial resolution are shown in the article. The features of planning high-resolution satellite imagery and remote sensing data processing require consideration of a larger number of additional influencing factors than when using traditional imagery with optical-electronic scanners characterized by sub-meter spatial resolution. The main stages of modeling and planning ultra-fast satellite imaging of lengthy territories, as well as the basic design formulas are considered in the article. To approximate a lengthy object defined by the nodal points on the map, smoothing with a cubic interpolating spline is used. For short forecast periods and low earth circular orbits, sufficient accuracy is achieved with the propagator SGP4, which allows to work with the initial conditions in TLE files generated by NORAD.

Thermal camouflage based on the phase-changing material GST

Camouflage technology has attracted growing interest for many thermal applications. Previous experimental demonstrations of thermal camouflage technology have not adequately explored the ability to continuously camouflage objects either at varying background temperatures or for wide observation angles. In this study, a thermal camouflage device incorporating the phase-changing material Ge2Sb2Te5 (GST) is experimentally demonstrated. It has been shown that near-perfect thermal camouflage can be continuously achieved for background temperatures ranging from 30 °C to 50 °C by tuning the emissivity of the device, which is attained by controlling the GST phase change. The thermal camouflage is robust when the observation angle is changed from 0° to 60°. This demonstration paves the way toward dynamic thermal emission control both within the scientific field and for practical applications in thermal information.

Calculation model of scattering depolarization for camouflaged target detection system

As a new physical model to identify the camouflaged targets in the complex environments, the main effect of the polarization measurement system is the scattering depolarization information of different measured parts. This paper had observed the multiple scattering theory of electromagnetic wave, and analyzed the mechanism of the relation between the depolarization and scattering path. The correctness and accuracy of the theoretical model had been verified by the Fresnel formula. In order to prove the practicability of this model, the experiments on two kinds of camouflage materials had been conducted at seven different angles. Research shows that, as a link between material physical characteristics and optical signal information, the model could accurately and effectively identify the moving camouflaged objects. Therefore, this method has a great application value, and the paper has very important significance on the development of tracking technology.