Detecting bottle caps defects
More than one million bottles are sold per minute. To fill, cap, and seal the bottles, manufacturers typically use high-speed assembly lines. It is important to verify that the caps are properly placed on the bottles. 100% inspection is crucial to ensure product safety and quality. Computer vision solutions perform defect detection with high accuracy and high speed.
Industries
Bottle cap defects detection is vital for many industries including beverage industry, pharmaceutical industry, cosmetics industry, food industry, and chemical industry. In such industries outflow of defective or faulty products is unacceptable and it is important to prevent spoilage, leakage, contamination and adulteration.
Benefits
Detecting most types of the bottle caps defects with high precision, computer vision solutions lead to significant benefits for producers and help to reduce production cost, shorten production cycle, ensure product quality, improve efficiency, speed up the inspection, enhance safety (especially in pharmaceutical and food industries where products must be safe to consume), reduce material waste and as a result reduce customer complaints, increase customer satisfaction and minimize the risks of damaging the brand's reputation.
The process of detecting bottle caps defects
Data Monsters provide high-performance computer vision solutions that can be integrated into production lines to inspect bottle caps. The solutions use cameras to capture high-resolution images of bottles as they move along the conveyor belt and machine learning algorithms to detect bottle caps defects.
The algorithm of detecting bottle cap defects includes the following steps:
- Image acquisition (industrial cameras take high-speed images of moving bottles of various shapes and designs and pass them to the pipeline);
- Pre-processing. This step includes noise removal, contrast enhancement, lighting conditions normalization and so on;
- Object detection and segmentation. The bottles are detected on the image and segmented;
- Feature extraction. Usually the shape, texture and color of bottle caps are extracted in this step;
- Defect detection. The extracted features are used to detect abnormalities. Texture inspection allows to detect surface defects like cracks, dents, and scratches. By comparing inspected cap shape against a standard shape, it is possible to identify deformations. Also, to detect defects the algorithm compares the color of the cap against a standard color.
To train the model machine learning algorithms use datasets of images of bottle caps with and without defects. The result is the classification of bottle caps into 2 categories: good and defective.
Using machine learning models in combination with image processing techniques can provide higher accuracy compared to relying solely on image processing, as the models can be trained to detect subtle patterns that may be difficult for humans to identify. The solutions show high accuracy even with different light and background conditions and are applicable for different types of bottle caps (plastic, rubber, aluminum, and even anti-theft caps).
Types of bottle caps defects that can be detected
Types of defects that can be detected using our solutions depend on the requirements of the application, the desired level of accuracy, speed, and cost and include:
- damages (chips, cracks, scratches);
- deformations (pinches, bents, dents, warping, buckling, roundness loss);
- missing parts (absence of caps, absence of tamper rings);
- loose caps, inclined caps, seated caps;
- defects of the sealing surface of the cap (not flat surface, not smooth surface);
- misaligned seals;
- double gaskets;
- contamination (dust, dirt, other particles, black spots);
- foreign objects (debris, insects, hair);
- voids.
Performance
The experimental results show that the proposed solution is 96.5% accurate in determining bottle caps defects.
Processing speed: 70,000..130,000 bottles / hour, depending on the hardware and resolution