Patent published on December 26, 2023

Robotic System's New Patent to Help Computers Recognize Objects and People in Pictures and Videos

In today's digital age, the ability of computers to understand and interpret visual information has become increasingly important. Computers are now expected to identify and categorize objects and people in pictures and videos accurately. This can be a challenging task due to the complexity of images and the various perspectives in which objects and people appear.

Fortunately, a new patent has been published by UBKANG (QINGDAO) TECHNOLOGY CO., offering a groundbreaking solution to this problem. The patent, titled "Human-object scene recognition method, device, and computer-readable storage medium" (patent number: US11854255B2), introduces a novel approach to help computers recognize objects and individuals in visual media.

The core problem addressed by this patent is that conventional image recognition methods primarily rely on low-level visual features, such as color, shape, and texture. While these features provide visual information, they fail to incorporate semantic understanding, resulting in significant positioning errors and a limited understanding of the scene depicted.

To overcome this limitation, the patented method utilizes special images called RGB and depth images. By analyzing these images and implementing complex algorithms, the computer can determine the position and direction of objects and people within the scene accurately. The integration of the depth information allows for a three-dimensional understanding of the objects and individuals, leading to improved accuracy and reliability.

The advantages of this method are significant. By outputting information regarding the proximity of objects or people to the target object, it enables effective guidance for human-robot and environment interaction. For instance, in robotics, this technology can facilitate target position setup and navigation through occlusion avoidance.

To further enhance accuracy, the patented method employs techniques like shrinkage contour and volume compensation to eliminate misclassification values. Additionally, the use of Convex Hull helps preserve essential information with minimal memory and CPU usage. In cases where occlusion causes segmentation challenges, stereo-based calculation methods merge the pieces into a coherent object.

Implemented as a robotic object recognition system, this invention has the potential to revolutionize various industries. In hospitals, it can assist in patient monitoring by recognizing medical equipment or identifying staff members. In factories, it can aid in quality control by identifying faulty products or irregularities on the assembly line. Similarly, in warehouses, malls, airports, and even homes, this technology can facilitate seamless automation and enhance security measures.

It is essential to note that, while this patent presents an innovative solution to a long-standing problem, there is no guarantee of its appearance in the market. It is ultimately up to UBKANG (QINGDAO) TECHNOLOGY CO. to determine the fate of this invention.

In conclusion, UBKANG's groundbreaking patent on a human-object scene recognition method brings exciting possibilities for computers to accurately understand and interpret visual information. With the potential to revolutionize various industries, the integration of this technology could streamline automation and enhance security measures while providing a more comprehensive understanding of the visual world we interact with.

P.S. Please note that the patent discussed in this article (patent number: US11854255B2) is currently in the patent stage, and there is no certainty regarding its future availability in the market.

Explore more