Robust and fast pedestrian detection method for far-infrared automotive driving assistance systems

Liu, Qiong; Zhuang, Jiajun; Ma, Jun

doi:10.1016/j.infrared.2013.06.003

Cited by 59 publications

(29 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use the LSI Far Infrared Pedestrian Dataset (LSIFIR) [18]. In the classification training set each example is a 64 × 32 image patch.…”

Section: Logistic Regression-based Classificationmentioning

confidence: 99%

Object Detection and Tracking

Yang

2016

Advanced Technologies in Modern Robotic Applications

View full text Add to dashboard Cite

Visual sensors provide comprehensive and abundant information of surrounding environment. In this chapter, we will first give a brief introduction of visual recognition with basic concepts and algorithms, followed by the introduction of the useful software toolkit JavaScript Object Notation (JSON) framework. Then we will review a series of vision-based object recognition and tracking techniques. The detection and classification of pedestrians in infrared thermal images is investigated using deep learning method. And the algorithm for tracking single moving objects based on JSON visual recognition framework is also introduced. As the extension of single moving objects tracking, we will also show the visual tracking to multiple moving objects with the aid of the particle swarm optimization (PSO) method. Introduction of Machine Vision RecognitionFrom our human's perspective, recognition is about searching and comparison. Our brain has a huge knowledge base, storing tens of thousands of objects. Sometimes images themselves are stored in the brain, in other cases only abstract features are stored. Those images and corresponding features are called references. Then, our eyes are opened, new images are transmitted to the brain. Those test images are compared with references, and conclusions come out.The visual recognition in machine is in some extent similar to humans' recognition. In order to recognize an object in a given image, machine needs to know what the object looks like. It learns the knowledge by loading some references before hand. Then, given a test image, matches between the references and the test samples are constructed. The machine then decides if there exists the wanted object in the test image, and locates it if possible, by analyzing the result of matches. Generally speaking, recognition is divided into two parts: detecting and locating. Detecting checks whether there exists a specific kind of object in the image. Then, locating marks regions of the detecting objects in the image.There are three key concepts that are helpful to understand the recognition problem: feature space, similarity metric, and search space and strategy. Feature space: Feature space determines the information used in matching. It may be the image itself, but specific features are used more often: edges, contours, surfaces, corners, etc. Intrinsic structures which imply the invariant properties of an image are preferred features.Similarity metric: The second key concept is related to the selection of a similarity metric. Since the metric measures the similarity among feature vectors, it is closely related to the selection of features. Typical similarity metrics include crosscorrelation, sum of absolute difference, Fourier phase correlation, sum of distance of nearest points, and many others. The choice of similarity metric is also thought to be an important task to determine the transformation between two images.Search space and strategy: The third key concept is search space and strategy. If the test image can be obtain...

show abstract

“…We use the LSI Far Infrared Pedestrian Dataset (LSIFIR) [18]. In the classification training set each example is a 64 × 32 image patch.…”

Section: Logistic Regression-based Classificationmentioning

confidence: 99%

Object Detection and Tracking

Yang

2016

Advanced Technologies in Modern Robotic Applications

View full text Add to dashboard Cite

show abstract

“…But recent studies have proved that Compared with visible images, thermal images are represented with different intensity maps, and not sensitive to illumination change and complicated details. Besides, thermal images can provide an enhanced spectral range that is imperceptible to human beings and contribute to obvious contrast between objects of high temperature variance and the environment (Zin et al, 2011;Liu et al, 2013;Bertozzi et al, 2007;Fang et al, 2004;.…”

Section: Comparisonmentioning

confidence: 99%

Pedestrians’ detection in thermal bands – Critical survey

Negied

Hemayed

Fayek

2015

Journal of Electrical Systems and Information Technology

View full text Add to dashboard Cite

Thermal imaging is simply the technique of using the heat given off by an object to produce an image of it or locate it. New thermal imaging frameworks for detection, segmentation and unique feature extraction and similarity measurements for human physiological biometrics recognition have been introduced in literature. The research investigates specialized algorithms that would use the individual's heat signature for human detection, crowd counting and applications that take benefits of this new technology. The highly accurate results obtained by the algorithms presented clearly demonstrate the ability of the thermal infrared systems to extend in application to other thermal imaging based systems.

show abstract

“…The histogram of oriented gradients (HOG) and its variants are widely used for pedestrian detection in IR images [7,5]. For example, Bin et al [5] proposed a novel descriptor called the scattered difference of directional gradients (SDDG), while Liu et al [8] proposed a pyramid entropy weighted HOG. The extracted features were then employed for binary classification using either template matching [9] or machine learning algorithms, such as the support vector machine (SVM) [7].…”

Section: Introduction and Related Workmentioning

confidence: 99%

Pedestrian detection in thermal images using adaptive fuzzy C-means clustering and convolutional neural networks

John

Mita

Zheng

et al. 2015

2015 14th IAPR International Conference on Machine Vision Applications (MVA)

View full text Add to dashboard Cite

Pedestrian detection is paramount for advanced driver assistance systems (ADAS) and autonomous driving. As a key technology in computer vision, it also finds many other applications, such as security and surveillance etc. Generally, pedestrian detection is conducted for images in visible spectrum, which are not suitable for night time detection. Infrared (IR) or thermal imaging is often adopted for night time due to its capability of capturing the emitted energy from pedestrians. The detection process firstly extracts candidate pedestrians from the captured IR image. Robust feature descriptors are formulated to represent those candidates. A binary classification of the extract features is then performed with trained classifier models. In this paper, an algorithm for pedestrian detection from IR image is proposed, where an adaptive fuzzy C-means clustering and convolutional neural networks are adopted. The adaptive fuzzy C-means clustering is used to segment the IR images and retrieve the candidate pedestrians. The candidate pedestrians are then pruned using human posture characteristics and the second central moments ellipse. The convolutional neural network is used to simultaneously learn relevant features and perform the binary classification. The performance of the proposed algorithm is compared with state-of-the-art algorithms on publicly available data set. A better detection accuracy with reduced computational accuracy is achieved.

show abstract

Robust and fast pedestrian detection method for far-infrared automotive driving assistance systems

Cited by 59 publications

References 32 publications

Object Detection and Tracking

Object Detection and Tracking

Pedestrians’ detection in thermal bands – Critical survey

Pedestrian detection in thermal images using adaptive fuzzy C-means clustering and convolutional neural networks

Contact Info

Product

Resources

About