SBIR Phase I: Assistive Digital Vision for the Blind

这个小型企业创新研究(SBIR)第一阶段项目旨在将最先进的图像处理算法实施到嵌入式设备上，该算法旨在识别当前需要高功率计算机处理资源的复杂背景中的对象，这将允许在小型、便携和负担得起的设备中使用这一功能。该项目还旨在包括一种避障算法，以识别特定范围内的物体。这项高风险的研究有望带来广泛的好处，重点是创造一种辅助数字视觉技术。这项工作的主要目标包括：1)将摄像机和声纳集成到商用嵌入式设备上；2)开发声纳传感器的障碍物检测/避障算法，并将其成功地实现到嵌入式设备中；3)定制摄像机传感器的对象识别/识别算法，以将其实现到嵌入式设备中。成功的评估将通过测量合成的面包板设备的性能来评估，该设备能够检测不同距离的物体，并在杂乱的背景下识别三个常见的物体。该项目的目标和预期结果是超过80%的检测概率。该项目更广泛的影响/商业潜力是使盲人能够对周围环境形成更全面的心理图景，并提高他们的情境意识。全世界有3900万视障人士。导盲犬和白色手杖是首选的行动辅助工具，但他们几乎没有其他既方便使用又负担得起的辅助工具。设想的技术将与白色手杖兼容，并将提醒用户地面上存在障碍物，如交通和标志杆以及悬挂物。此外，物体识别将进一步允许用户对其周围环境形成更全面的心理图景，并提高其移动性。虽然其他导航支持产品的价格高达数百至数千美元，但本文提出的技术预计将以不到200美元的价格商业化，从而使其能够对广大视障人士产生广泛影响。在便携式设备中包含嵌入式摄像头和声纳技术的合成技术也可能有助于机器人和人工智能领域，在这些领域，对环境的更好理解将为新的决策算法提供支持。

This Small Business Innovation Research (SBIR) Phase I project seeks to implement state-of-the art image processing algorithms designed to identify objects amidst a complex background, which currently require high power computer processing resources, onto an embedded device, which would allow the capability to be used in small, portable, and affordable devices. The project further aims to include an obstacle avoidance algorithm to identify objects within a specific range. This high-risk research promises wide-ranging benefits, with the focus to be on creation of an assistive digital vision technology. The main objectives of the proposed effort include: 1) integrating a camera and sonar onto a commercially available embedded device; 2) development of an obstacle detection/avoidance algorithm for the sonar sensor and successfully implementing it into the embedded device; 3) customization of object identification / recognition algorithms for the camera sensor to implement them into the embedded device. Success will be evaluated by measuring performance of the resultant breadboard device in its ability to detect objects at various distances, and to recognize three common objects against a cluttered background. The goal and expected result is to exceed a probability of detection of 80%.The broader impact/commercial potential of this project is enabling blind people to develop a more comprehensive mental picture of their surrounding and improving their situational awareness. There are 39 million visually impaired people living around the world. Guide dogs and white canes are the preferred mobility aids, but there is little else available to them that is both user-friendly and affordable. The envisaged technology will be compatible with the white cane and will alert the user to the presence of above ground obstacles, such as traffic and sign poles and overhanging objects. In addition, the object identification will further allow the user to develop a more comprehensive mental picture of his surroundings and improve his mobility. While other navigation support products are in the high hundreds to thousands of dollars, the technology proposed herein is expected to be commercialized in a product available for less than $200, thus allowing it to make a broad impact to the wide population of the visually impaired. The resultant technology that contains embedded camera and sonar technologies in a portable device may also contribute to the field of robotics and artificial intelligence, wherein a better understanding of the environment will support new decision-making algorithms.