On-line learning of mutually orthogonal subspaces for face recognition by image sets.

We address the problem of face recognition by matching image sets. Each set of face images is represented by a subspace (or linear manifold) and recognition is carried out by subspace-to-subspace matching. In this paper, 1) a new discriminative method that maximises orthogonality between subspaces is proposed. The method improves the discrimination power of the subspace angle based face recognition method by maximizing the angles between different classes. 2) We propose a method for on-line updating the discriminative subspaces as a mechanism for continuously improving recognition accuracy. 3) A further enhancement called locally orthogonal subspace method is presented to maximise the orthogonality between competing classes. Experiments using 700 face image sets have shown that the proposed method outperforms relevant prior art and effectively boosts its accuracy by online learning. It is shown that the method for online learning delivers the same solution as the batch computation at far lower computational cost and the locally orthogonal method exhibits improved accuracy. We also demonstrate the merit of the proposed face recognition method on portal scenarios of multiple biometric grand challenge.

Multiscale analysis of factors that affect the distribution of sharks throughout the northern Gulf of Mexico

Identification of the spatial scale at which marine communities are organized is critical to proper management, yet this is particularly difficult to determine for highly migratory species like sharks. We used shark catch data collected during 2006–09 from fishery-independent bottom-longline surveys, as well as biotic and abiotic explanatory data to identify the factors that affect the distribution of coastal sharks at 2 spatial scales in the northern Gulf of Mexico. Centered principal component analyses (PCAs) were used to visualize the patterns that characterize shark distributions at small (Alabama and Mississippi coast) and large (northern Gulf of Mexico) spatial scales. Environmental data on temperature, salinity, dissolved oxygen (DO), depth, fish and crustacean biomass, and chlorophyll-a (chl-a) concentration were analyzed with normed PCAs at both spatial scales. The relationships between values of shark catch per unit of effort (CPUE) and environmental factors were then analyzed at each scale with co-inertia analysis (COIA). Results from COIA indicated that the degree of agreement between the structure of the environmental and shark data sets was relatively higher at the small spatial scale than at the large one. CPUE of Blacktip Shark (Carcharhinus limbatus) was related positively with crustacean biomass at both spatial scales. Similarly, CPUE of Atlantic Sharpnose Shark (Rhizoprionodon terraenovae) was related positively with chl-a concentration and negatively with DO at both spatial scales. Conversely, distribution of Blacknose Shark (C. acronotus) displayed a contrasting relationship with depth at the 2 scales considered. Our results indicate that the factors influencing the distribution of sharks in the northern Gulf of Mexico are species specific but generally transcend the spatial boundaries used in our analyses.

MDNR-NOAA trawl standardization study

Long-term living resource monitoring programs are commonly conducted globally to evaluate trends and impacts of environmental change and management actions. For example, the Woods Hole bottom trawl survey has been conducted since 1963 providing critical information on the biology and distribution of finfish and shellfish in the North Atlantic (Despres-Patango et al. 1988). Similarly in the Chesapeake Bay, the Maryland Department of Natural Resources (MDNR) Summer Blue Crab Trawl survey has been conducted continuously since 1977 providing management-relevant information on the abundance of this important commercial and recreational species. A key component of monitoring program design is standardization of methods over time to allow for a continuous, unbiased data set. However, complete standardization is not always possible where multiple vessels, captains, and crews are required to cover large geographic areas (Tyson et al. 2006). Of equal issue is technological advancement of gear which serves to increase capture efficiency or ease of use. Thus, to maintain consistency and facilitate interpretation of reported data in long-term datasets, it is imperative to understand and quantify the impacts of changes in gear and vessels on catch per unit of effort (CPUE). While vessel changes are inevitable due to ageing fleets and other factors, gear changes often reflect a decision to exploit technological advances. A prime example of this is the otter trawl, a common tool for fisheries monitoring and research worldwide. Historically, trawl nets were constructed of natural materials such as cotton and linen. However modern net construction consists of synthetic materials such as polyamide, polyester, polyethylene, and polypropylene (Nielson et. al. 1983). Over the past several decades, polyamide materials which will be referred to as nylon, has been a standard material used in otter trawl construction. These trawls are typically dipped into a latex coating for increased abrasion resistance, a process that is referred to as “green dipped.” More recently, polyethylene netting has become popular among living resource monitoring agencies. Polyethylene netting, commonly known as sapphire netting, consists of braided filaments that form a very durable material more resistant to abrasion than nylon. Additionally, sapphire netting allows for stronger knot strength during construction of the net further increasing the net’s durability and longevity. Also, sapphire absorbs less water with a specific gravity near 0.91 allowing the material to float as compared to nylon with specific gravity of 1.14 (Nielson et. al. 1983). This same property results in a light weight net which is more efficient in deployment, retrieval and fishing of the net, particularly when towing from small vessels. While there are many advantages to the sapphire netting, no comparative efficiency data is available for these two trawl net types. Traditional nylon netting has been used consistently for decades by the MDDNR to generate long term living resource data sets of great value. However, there is much interest in switching to the advanced materials. In addition, recent collaborative efforts between MDNR and NOAA’s Cooperative Oxford Laboratory (NOAA-COL) require using different vessels for trawling in support of joint projects. In order to continue collaborative programs, or change to more innovative netting materials, the influence of these changes must be demonstrated to be negligible or correction factors determined. Thus, the objective of this study was to examine the influence of trawl net type, vessel type, and their interaction on capture efficiency.