Computer vision hardware and image analysis tools for automated inspection in the food processing industry

用于食品加工行业自动检测的计算机视觉硬件和图像分析工具

• 批准号: 138061-2007
• 负责人: Krouglicof, Nicholas
• 金额: $ 1.09万
• 依托单位: Memorial University of Newfoundland
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2010
• 资助国家: 加拿大
• 起止时间: 2010-01-01 至 2011-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=451751
• 关键词: Computer; vision; hardware; image; analysis

The growing demand for fast, accurate and reliable quality inspection tools for the food processing industry has been driven by a number of factors; more stringent government safety standards, rising labor costs, and increased consumer expectations for higher quality food products. In many food processing applications, human inspectors are simply too subjective and unreliable given the time allotted to perform the inspection task and the tedious nature of the work. For these reasons, computer vision is proving to be a rapid, objective and cost effective alternative to human inspection for a growing number of food products. However, significant technological challenges remain particularly with respect to the inspection of "natural foods" such as fish and poultry. The proposed research program addresses these challenges though the development of both computer vision hardware and image analysis tools specifically for the automated inspection of natural foods. The ultimate goal is the development of a rapid, objective, robust and cost effective automated inspection system that improves food quality and consumer safety. The proposed research program is based on the hypothesis that no single detection technology will ever adequately satisfy the requirements of the food processing industry. A more promising approach is to base the inspection on multiple images acquired using various regions of the EM spectrum (visible, near-infrared, ultraviolet, X-ray) combined with different illumination strategies (reflection, trans-illumination, fluorescence). This approach (i.e., real-time multispectral image fusion) can potentially provide the basis for an accurate and robust automated food inspection system. In order to achieve real-time performance, research will focus on implementing high level image processing functions in hardware with the long-term goal of developing an "intelligent" camera for automated inspection that incorporates an entire vision system within a single compact enclosure. The development of such advanced technology for food inspection is critical to Canada both for addressing increasingly sophisticated consumer demands and ensuring food safety and competitiveness in an increasingly global marketplace.

食品加工行业对快速、准确和可靠的质量检测工具的需求不断增长，这是由许多因素推动的;更严格的政府安全标准、不断上升的劳动力成本以及消费者对更高质量食品的期望。 在许多食品加工应用中，考虑到分配给执行检查任务的时间和工作的繁琐性质，人类检查员过于主观和不可靠。由于这些原因，计算机视觉被证明是一种快速，客观和具有成本效益的替代人工检测越来越多的食品。然而，仍然存在重大的技术挑战，特别是在检查鱼和家禽等“天然食品”方面。拟议的研究计划通过开发专门用于天然食品自动检测的计算机视觉硬件和图像分析工具来解决这些挑战。最终目标是开发一种快速、客观、强大且具有成本效益的自动化检测系统，以提高食品质量和消费者安全。拟议的研究计划是基于这样的假设，即没有一种检测技术能够充分满足食品加工行业的要求。 一种更有前途的方法是将检查基于使用EM光谱（可见光、近红外、紫外、X射线）的各个区域与不同照明策略（反射、透射照明、荧光）相结合获得的多个图像。这种方法（即，实时多光谱图像融合）可以潜在地为精确和鲁棒的自动食品检查系统提供基础。为了实现实时性能，研究将集中在硬件中实现高级图像处理功能，长期目标是开发一种用于自动检测的“智能”相机，该相机将整个视觉系统集成在一个紧凑的外壳中。开发这种先进的食品检验技术对加拿大来说至关重要，因为它既能满足日益复杂的消费者需求，又能确保食品安全和在日益全球化的市场中的竞争力。