Orientation (computer vision)

About: Orientation (computer vision) is a research topic. Over the lifetime, 17196 publications have been published within this topic receiving 358181 citations.

Line detection in images through regularized Hough transform

Abstract: The problem of determining the location and orientation of straight lines in images is often encountered in the fields of computer vision and image processing. Traditionally the Hough transform has been widely used to solve this problem for binary images, due to it's simplicity and effectiveness. In this paper we pose the problem of detecting straight lines in gray-scale images as an inverse problem. We treat the input image as noisy observations, which are related to the underlying transform domain image through the inverse Hough transform operator. We then regularize this inverse problem using constraints that accentuate peaks in the Hough parameter space. We present four different forms of such constraints and demonstrate their effectiveness. Finally we show how our scheme can be alternatively viewed as one of finding an optimal representation of the image in terms of elements chosen from a redundant dictionary of lines, and thus is a form of adaptive signal representation.

Learning Sparse Topographic Representations with Products of Student-t Distributions

Abstract: We propose a model for natural images in which the probability of an image is proportional to the product of the probabilities of some filter outputs. We encourage the system to find sparse features by using a Student-t distribution to model each filter output. If the t-distribution is used to model the combined outputs of sets of neurally adjacent filters, the system learns a topographic map in which the orientation, spatial frequency and location of the filters change smoothly across the map. Even though maximum likelihood learning is intractable in our model, the product form allows a relatively efficient learning procedure that works well even for highly overcomplete sets of filters. Once the model has been learned it can be used as a prior to derive the "iterated Wiener filter" for the purpose of denoising images.

Automatic three-dimensional inspection of patient setup in radiation therapy using portal images, simulator images, and computed tomography data

Abstract: In external beam radiotherapy, conventional analysis of portal images in two dimensions (2D) is limited to verification of in-plane rotations and translations of the patient. We developed and clinically tested a new method for automatic quantification of the patient setup in three dimensions (3D) using one set of computed tomography (CT) data and two transmission images. These transmission images can be either a pair of simulator images or a pair of portal images. Our procedure adjusts the position and orientation of the CT data in order to maximize the distance through bone in the CT data along lines between the focus of the irradiation unit and bony structures in the transmission images. For this purpose, bony features are either automatically detected or manually delineated in the transmission images. The performance of the method was quantified by aligning randomly displaced CT data with transmission images simulated from digitally reconstructed radiographs. In addition, the clinical performance were assessed in a limited number of images of prostate cancer and parotid gland tumor treatments. The complete procedure takes less than 2 min on a 90-MHz Pentium PC. The alignment time is 50 s for portal images and 80 s for simulator images. The accuracy is about 1 mm and 1 degrees. Application to clinical cases demonstrated that the procedure provides essential information for the correction of setup errors in case of large rotations (typically larger than 2 degrees) in the setup. The 3D procedure was found to be robust for imperfections in the delineation of bony structures in the transmission images. Visual verification of the results remains, however, necessary. It can be concluded that our strategy for automatic analysis of patient setup in 3D is accurate and robust. The procedure is relatively fast and reduces the human workload compared with existing techniques for the quantification of patient setup in 3D. In addition, the procedure improves the accuracy of treatment verification in 2D in some cases where rotational deviations in the setup occur.

Gyro in the air: tracking 3D orientation of batteryless internet-of-things

Abstract: 3D orientation tracking is an essential ingredient for many Internet-of-Things applications. Yet existing orientation tracking systems commonly require motion sensors that are only available on battery-powered devices. In this paper, we propose Tagyro, which attaches an array of passive RFID tags as orientation sensors on everyday objects. Tagyro uses a closed-form model to transform the run-time phase offsets between tags into orientation angle. To enable orientation tracking in 3D space, we found the key challenge lies in the imperfect radiation pattern of practical tags, caused by the antenna polarity, non-isotropic emission and electromagnetic coupling, which substantially distort phase measurement. We address these challenges by designing a set of phase sampling and recovery algorithms, which together enable reliable orientation sensing with 3 degrees of freedom. We have implemented a real-time version of Tagyro on a commodity RFID system. Our experiments show that Tagyro can track the 3D orientation of passive objects with a small error of 4°, at a processing rate of 37.7 samples per second.

Extracting road curvature and orientation from image edge points without perceptual grouping into features

Abstract: The ARCADE (Automated Road Curvature And Direction Estimation) algorithm estimates road curvature and tangential road orientation relative to the camera line-of-sight. The input to ARCADE consists of edge point locations and orientations extracted from an image, and it performs the estimation without the need for any prior perceptual grouping of the edge points into individual lane boundaries. It is able to achieve this through the use of global constraints on the individual lane boundary shapes derived from an explicit model of road structure in the world. The use of the least median squares robust estimation technique allows the algorithm to function correctly in cases where up to 50% of the input edge data points are contaminating noise. Two applications of ARCADE as the first stage of processing for a lane sensing task are described: 1) the extraction of the locations of the features defining the visible lane structure of the road; and 2) the generation of training instances for an ALVINN-like neural network road follower.