Review on Visual Characteristic Measurement Research of Aquatic Animals Based on Computer Vision
Duan Yane,Li Daoliang,Li Zhenbo,Fu Zetian
DOI: https://doi.org/10.11975/j.issn.1002-6819.2015.15.001
2015-01-01
Abstract:In aquaculture, visual attribute information of aquatic animals is the basis of determining growth condition, feed conversion, medication dosage, harvesting date and grading for aquaculture farmers and managers. For improving the quality of aquatic products, the automatic and non-destructive measurement of visual attributes is becoming more and more important in modern fishery. For decades, computer vision, as a non-destructive, rapid, economic, consistent, reliable and objective inspection tool based on image analysis and processing with a variety of applications, has been gradually used in visual quality detection of aquatic animals. Quite a number of researches have highlighted its potential application in aquaculture. Underwater or overwater video/image measurement systems based on image processing technologies have been used widely for automatically counting and measuring fish in aquaculture, fisheries and conservation management. However, the application of computer vision technologies in aquaculture is very challenging because the inspected objects are sensitive, easily stressed and free to move in an environment in which lighting, visibility and stability are generally not controllable, and the camera must be operated underwater or in a wet environment. This review updates and summarizes recent representative researches and industrial solutions proposed in order to evaluate the general trends of computer vision and image processing in the visible range applied for inspection of aquatic animals. On the basis of introducing the mode of operation and the components of a computer vision detection system, this paper presents a review of the overseas and domestic research status in visual attribute measurement of aquatic animals according to inspection tasks that are common to almost all visual attribute detection systems of aquatic animal: measurement of size and shape parameters, estimation of mass and quantification of color, etc. Specially, the techniques involve in computer vision detection system used for the improvements of data acquisition environment, accuracy of object detection and contour extraction, and the measuring results are analyzed in detail, including the consideration of image acquisition method and mode, the development of fish detection and feature points definition algorithm, as well as the study about feature computation method and mass prediction model. In addition, the comprehensive application of computer vision detection technology in aquatic animals is also discussed, including disease diagnosis, identification of species, detection of gender and age, as well as grading and sorting of fish. The objective of the review is to highlight the development of computer vision systems, image analysis and processing approaches in aquaculture and analyze the advantages and limitations of current computer vision detection systems which have made some progresses, but have not matured into a useful tool in aquaculture. Considering the overall trends, we propose some future research directions of the computer vision detection systems for aquatic animals, including the technology of image acquisition in natural underwater environment, complete process of fish detection and contour extraction, seamless integration of modules, as well as the technology of multi-information fusion. With the future development in these areas, computer vision detection technique may achieve higher accuracy and efficiency, and wider application in visual quality detection of aquatic animals.