Yakov Diskin

Automatic segmentation of carcinoma in radiographs

A strong emphasis has been made on making the healthcare system and the diagnostic procedure more efficient. In this paper, we present an automatic detection technique designed to segment out abnormalities in X-ray imagery. Utilizing the proposed algorithm allows radiologists and their assistants to more effectively sort and analyze large amount of imagery. In radiology, X-ray beams are used to detect various densities within a tissue and to display accompanying anatomical and architectural distortion. Lesion localization within fibrous or dense tissue is complicated by a lack of clear visualization as compared to tissues with an increased fat distribution. As a result, carcinoma and its associated unique patterns can often be overlooked within dense tissue. We introduce a new segmentation technique that integrates prior knowledge, such as intensity level, color distribution, texture, gradient, and shape of the region of interest taken from prior data, within segmentation framework to enhance performance of region and boundary extraction of defected tissue regions in medical imagery. Prior knowledge of the intensity of the region of interest can be extremely helpful in guiding the segmentation process, especially when the carcinoma boundaries are not well defined and when the image contains non-homogeneous intensity variations. We evaluate our algorithm by comparing our detection results to the results of the manually segmented regions of interest. Through metrics, we also illustrate the effectiveness and accuracy of the algorithm in improving the diagnostic efficiency for medical experts.

Dense 3D point-cloud model using optical flow for a monocular reconstruction system

In this paper, we present an enhanced 3D reconstruction algorithm designed to support an autonomously navigated unmanned ground vehicle. An unmanned system can use the technique to construct a point cloud model of its unknown surroundings. The algorithm presented focuses on the 3D reconstruction of a scene using image sequences captured by only a single moving camera. The original reconstruction process, resulting with a point cloud, was computed utilizing extracted and matched Speeded Up Robust Feature (SURF) points from subsequent video frames. Using depth triangulation analysis, we were able to compute the depth of each feature point within the scene. We concluded that although SURF points are accurate and extremely distinctive, the number of points extracted and matched was not sufficient for our applications. A sparse point cloud model hinders the ability to do further processing for the autonomous system such as object recognition or self-positioning. We present an enhanced version of the algorithm which increases the number of points within the model while maintaining the near real-time computational speeds and accuracy of the original sparse reconstruction. We do so by generating points using both global image characteristics and local SURF feature neighborhood information. Specifically, we generate optical flow disparities using the Horn-Schunck optical flow estimation technique and evaluate the quality of these features for disparity calculations using the SURF keypoint detection method. Areas of the image that locate within SURF feature neighborhoods are tracked using optical flow and used to compute an extremely dense model. The enhanced model contains the high frequency details of the scene that allow for 3D object recognition. The main contribution of the newly added preprocessing steps is measured by evaluating the density and accuracy of the reconstructed point cloud model in relation to real-world measurements.

Multi-Pose Face Recognition And Tracking System

Abstract We propose a real time system for person detection, recognition and tracking using frontal and profile faces. The system integrates face detection, face recognition and tracking techniques. The face detection algorithm uses both frontal face and profile face detectors by extracting the Haar’ features and uses them in a cascade of boosted classifiers. The pose is determined from the face detection algorithm which uses a combination of profile and frontal face cascades and, depending on the pose, the face is compared with a particular set of faces having the same range for classification. The detected faces are recognized by projecting them onto the Eigenspace obtained from the training phase using modular weighted PCA and then, are tracked using the Kalman filter multiple face tracker. In this proposed system, the pose range is divided into three bins onto which the faces are sorted and each bin is trained separately to have its own Eigenspace. This system has the advantage of recognizing and tracking an individual with minimum false positives due to pose variations.

Dense point-cloud representation of a scene using monocular vision

Abstract. We present a three-dimensional (3-D) reconstruction system designed to support various autonomous navigation applications. The system presented focuses on the 3-D reconstruction of a scene using only a single moving camera. Utilizing video frames captured at different points in time allows us to determine the depths of a scene. In this way, the system can be used to construct a point-cloud model of its unknown surroundings. We present the step-by-step methodology and analysis used in developing the 3-D reconstruction technique. We present a reconstruction framework that generates a primitive point cloud, which is computed based on feature matching and depth triangulation analysis. To populate the reconstruction, we utilized optical flow features to create an extremely dense representation model. With the third algorithmic modification, we introduce the addition of the preprocessing step of nonlinear single-image super resolution. With this addition, the depth accuracy of the point cloud, which relies on precise disparity measurement, has significantly increased. Our final contribution is an additional postprocessing step designed to filter noise points and mismatched features unveiling the complete dense point-cloud representation (DPR) technique. We measure the success of DPR by evaluating the visual appeal, density, accuracy, and computational expense and compare with two state-of-the-art techniques.

Vision-based navigation system for obstacle avoidance in complex environments

We present a mobile system capable of autonomous navigation through complex unknown environments that contain stationary obstacles and moving targets. The intelligent system is composed of several fine-tuned computer vision algorithms running onboard in real-time. The first of these utilizes onboard cameras to allow for stereoscopic estimation of depths within the surrounding environment. The novelty of the approach lies in algorithmic efficiency and the ability of the system to complete a given task through the utilization of scene reconstruction and in making real-time automated decisions. Secondly, the system performs human body detection and recognition using advanced local binary pattern (LBP) descriptors. The LBP descriptors allow the system to perform human identification and tracking tasks irrespective of lighting conditions. Lastly, face detection and recognition allow for an additional layer of biometrics to ensure the correct target is being tracked. The face detection algorithm utilizes the Voila-Jones cascades, which are combined to create a pose invariant face detection system. Furthermore, we utilize a modular principal component analysis technique to perform pose-invariant face recognition. In this paper, we present the results of a series of experiments designed to automate the security patrol process. Our mobile security system completes a series of tasks within varying scenarios that range in difficulty. The tasks consist of tracking an object in an open environment, following a person of interest through a crowded environment, and following a person who disappears around a corner.

Dense point-cloud creation using superresolution for a monocular 3D reconstruction system

We present an enhanced 3D reconstruction algorithm designed to support an autonomously navigated unmanned aerial system (UAS). The algorithm presented focuses on the 3D reconstruction of a scene using only a single moving camera. In this way, the system can be used to construct a point cloud model of its unknown surroundings. The original reconstruction process, resulting with a point cloud was computed based on feature matching and depth triangulation analysis. Although dense, this original model was hindered due to its low disparity resolution. As feature points were matched from frame to frame, the resolution of the input images and the discrete nature of disparities limited the depth computations within a scene. With the recent addition of the preprocessing steps of nonlinear super resolution, the accuracy of the point cloud which relies on precise disparity measurement has significantly increased. Using a pixel by pixel approach, the super resolution technique computes the phase congruency of each pixels neighborhood and produces nonlinearly interpolated high resolution input frames. Thus, a feature point travels a more precise discrete disparity. Also, the quantity of points within the 3D point cloud model is significantly increased since the number of features is directly proportional to the resolution and high frequencies of the input image. The contribution of the newly added preprocessing steps is measured by evaluating the density and accuracy of the reconstructed point cloud for autonomous navigation and mapping tasks within unknown environments.

Multi-modal low cost mobile indoor surveillance system on the Robust Artificial Intelligence-based Defense Electro Robot (RAIDER)

We present an autonomous system capable of performing security check routines. The surveillance machine, the Clearpath Husky robotic platform, is equipped with three IP cameras with different orientations for the surveillance tasks of face recognition, human activity recognition, autonomous navigation and 3D reconstruction of its environment. Combining the computer vision algorithms onto a robotic machine has given birth to the Robust Artificial Intelligencebased Defense Electro-Robot (RAIDER). The end purpose of the RAIDER is to conduct a patrolling routine on a single floor of a building several times a day. As the RAIDER travels down the corridors off-line algorithms use two of the RAIDERs side mounted cameras to perform a 3D reconstruction from monocular vision technique that updates a 3D model to the most current state of the indoor environment. Using frames from the front mounted camera, positioned at the human eye level, the system performs face recognition with real time training of unknown subjects. Human activity recognition algorithm will also be implemented in which each detected person is assigned to a set of action classes picked to classify ordinary and harmful student activities in a hallway setting.The system is designed to detect changes and irregularities within an environment as well as familiarize with regular faces and actions to distinguish potentially dangerous behavior. In this paper, we present the various algorithms and their modifications which when implemented on the RAIDER serves the purpose of indoor surveillance.

Adaptive noise suppression technique for dense 3D point cloud reconstructions from monocular vision

Mobile vision-based autonomous vehicles use video frames from multiple angles to construct a 3D model of their environment. In this paper, we present a post-processing adaptive noise suppression technique to enhance the quality of the computed 3D model. Our near real-time reconstruction algorithm uses each pair of frames to compute the disparities of tracked feature points to translate the distance a feature has traveled within the frame in pixels into real world depth values. As a result these tracked feature points are plotted to form a dense and colorful point cloud. Due to the inevitable small vibrations in the camera and the mismatches within the feature tracking algorithm, the point cloud model contains a significant amount of misplaced points appearing as noise. The proposed noise suppression technique utilizes the spatial information of each point to unify points of similar texture and color into objects while simultaneously removing noise dissociated with any nearby objects. The noise filter combines all the points of similar depth into 2D layers throughout the point cloud model. By applying erosion and dilation techniques we are able to eliminate the unwanted floating points while retaining points of larger objects. To reverse the compression process, we transform the 2D layer back into the 3D model allowing points to return to their original position without the attached noise components. We evaluate the resulting noiseless point cloud by utilizing an unmanned ground vehicle to perform obstacle avoidance tasks. The contribution of the noise suppression technique is measured by evaluating the accuracy of the 3D reconstruction.

3D Object Classification in Uncalibrated Structure from Motion Models

We present a 3D object classification technique that is robust to noise and inaccuracies of 3D Structure from Motion models. We illustrate our classification results on 3D models reconstructed from real-world aerial imagery.

3D Scene Reconstruction for Aiding Unmanned Vehicle Navigation

We present a 3D reconstruction algorithm designed to support various autonomous vehicle navigation applications. The algorithm presented focuses on the 3D reconstruction of a scene using only a single moving camera. Utilizing video frames captured at different points in time allows us to determine the relative depths in a scene. The original reconstruction process resulting in a point cloud was computed based on feature matching and depth triangulation analysis. In an improved version of the algorithm, we utilized optical flow features to create an extremely dense representation model. Although dense, this model is hindered due to its low disparity resolution. With the third algorithmic modification, we introduce the addition of the preprocessing step of nonlinear super resolution. With this addition, the accuracy and quantity of features is significantly increased since the number of features is directly proportional to the resolution and high frequencies of the input images. Our final contribution of additional pre and post processing steps are designed to filter noise points and mismatched features, completing the presentation of our Dense Point-cloud Representation (DPR) technique. We measure the success of DPR by evaluating the visual appeal, density, usability and computational expense of the reconstruction technique and compare with two state-of-the-art techniques.