DaoAI 3D Robot Vision for Food Foreign Object Removal and Quality Grading

In the food/agriculture industry, foreign object removal and quality grading are crucial steps to ensure product quality and safety. DaoAI 3D robot vision from WeLinkirt provides an efficient and accurate solution for this industry.

User Scenario: On the production line of a leading food manufacturer, the main products are various types of bagged snacks. The detection objects include foreign objects in the food, such as metal fragments and plastic particles, as well as the quality of the products, such as appearance integrity and size specifications. Before the snack packaging process, a comprehensive inspection of the products is required to ensure that the products entering the packaging process meet the quality standards.

Pain Points: In the traditional detection methods, manual inspection is inefficient, and there are prone to missed detections and false alarms. According to statistics, the missed detection rate of manual inspection is about 5%, and the false alarm rate reaches 10%. This not only increases the unqualified rate of products but also leads to a large waste of labor costs. At the same time, with the change of market demand, product model changes are frequent. The traditional detection equipment takes more than 30 minutes for model change, which seriously affects the production efficiency. Combining with today's hot-spot direction, traditional detection methods lack the learning ability of robot vision and embodied intelligence and cannot improve the detection ability by learning a large number of human work records.

Technical Principle

DaoAI 3D robot vision uses a self-developed 3D camera for imaging. This camera can obtain the three-dimensional shape information of objects and accurately determine the position and posture of objects in combination with the 6D pose estimation algorithm. For foreign object removal, it uses the feature recognition ability of the visual basic model to identify the foreign object features in the food through learning a large number of foreign object samples. In terms of quality grading, through the three-dimensional shape reconstruction technology, the appearance and size of the product are accurately measured to judge whether the product meets the quality standard. This technology is effective because the 3D camera provides more abundant information. Compared with traditional 2D imaging, it can more accurately identify foreign objects and judge product quality, and the 6D pose estimation provides accurate positioning for the robot's operation.

• The self-developed 3D camera obtains three-dimensional information to improve the detection accuracy.
• The 6D pose estimation algorithm realizes accurate object positioning.
• The visual basic model learns foreign object features to enhance the recognition ability.
• The three-dimensional shape reconstruction technology accurately judges product quality.

WeLinkirt Solution and Product

Centered on DaoAI 3D robot vision, a brain-eye - body closed-loop system is constructed. The self-developed 3D camera serves as the “eye” to obtain the three-dimensional information of the product in real-time; the 6D pose estimation algorithm and the visual basic model serve as the “brain” to analyze and judge the obtained information; the robot serves as the “body” to perform foreign object removal and quality grading operations according to the instructions of the “brain”. In actual implementation, it is deployed through SDK/API/Docker, supporting 100% local privatization to ensure that the data does not leave the factory. At the same time, it can be combined with the DaoAI AI AOI software system for positive sample/few-sample learning to improve the detection efficiency.

DaoAI 3D robot vision realizes an efficient closed-loop from data acquisition to decision-making and execution, providing reliable guarantee for foreign object removal and quality grading in the food industry.

Quantitative Results: After using DaoAI 3D robot vision, the foreign object detection rate reaches 98%, the missed detection rate is reduced to <2%, and the false alarm rate is reduced by -70%. The product model change time is shortened to 5min, greatly improving the production efficiency. At the same time, due to the improvement of detection accuracy, the product quality is effectively improved, reducing rework and customer complaints caused by quality problems.