Handling Adverse Visual Conditions for Target Tracking and Recognition
Abstract
Extending visual sensing capacity in target tracking and identification is very important for Army s future forces. Contemporary computer vision techniques assume mild visual conditions and depend on good quality imagery. In practice, however, as the environments are unconstrained, their performances are largely jeopardized by adverse visual conditions, e.g., induced by bad weather conditions, when the visual quality of the data is seriously degraded and the visual details may be obscured in such perceptually "low-quality" imagery. Most existing solutions to this challenge are to perform pre-processing that restores the quality of the imagery, e.g., via image super-resolution or de-blurring. However, such image restoration tasks themselves are very difficult and computationally demanding. Therefore, this solution is not practical. Except some ad hoc methods, satisfactory solutions are still yet to be found, which has impeded the development of "all-weather" vision systems. The goal of this project is to find an innovative solution to overcome this challenge, by exploring a unified approach that does not perform explicit image restoration as pre-processing in target tracking and recognition under various adverse visual conditions and degradations. Specifically, we plan to address the following issues: (1) A principled approach and its theoretical foundation. The key idea of avoiding performing explicit image restoration is to embed the prior knowledge for restoration into visual target matching. Our objective is to develop a general approach of learning image similarity and visual regression that applies to various situations of adverse visual conditions. (2) Visual matching and tracking. The performance of target tracking is largely determined by the quality of visual similarity metric and visual matching. It should be adaptive to different imagery. Our objective is to learn the metric for the low-quality imagery under adverse visual conditions, by steering and aligning the known metric from good-quality images. (3) Visual attribute estimation and identification. A target has various semantic descriptive attributes. Some are discrete and others are continuous. It is very difficult to extract and recognize them from perceptually low-quality images. Our objective is to estimate the attributes directly from low-quality images via learning the reconstruction-based visual regression. (4) Tools and prototype systems. We will develop effective and efficient tools for visual matching and visual attribute estimation that can be widely used in many computer vision tasks. In addition, we will develop prototype systems for visual target tracking and identification that handles various adverse visual conditions. One innovation of the proposed approach is that it avoids performing explicit, expensive and dedicated image/video restoration, but rather using the low-quality data directly. This new research will lead to innovative and computationally efficient solutions to handle adverse visual conditions. Moreover, it is a principled and general solution that will be able to handle various image degradations in the same framework. It will empower effective visual matching and visual regression. This will make possible target matching, tracking and recognition on low-quality data in various adverse visual conditions. In addition, image restoration can also be done in the proposed new approach as by-products, and the same approach also provides a general solution to target attribute estimation and recognition, which greatly benefits target identification and understanding. This research will have significant impacts on the Army by making possible "all-weather" computer vision systems for target tracking and recognition in a wide range of application scenarios, including video surveillance, Digital Soldiers, military robots, and intelligence gathering for urban warfare, etc..
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1610138
Entities
People
- Ying Wu
Organizations
- Army Contracting Command
- Northwestern University
- United States Army