Visual Solution for Plastic Bottle Appearance Inspection

Project Requirements

As the picture shows:

1. Detect scratches on the surface of plastic bottles.

2. Detect the black spots inside the plastic bottle.

3. Detect plastic bottle dust.

4. Detect the internal stripes of plastic bottles.

5. Accuracy requirements: not provided.

6. CT requirements: not provided.

Scheme Process

Scheme selection:

Using a 4K line scan camera, the field of view is 60mm, and the single-pixel accuracy is 0.0146mm / pix. The maximum line scan speed is 5.9568m / min.

Action flow: fixed shot

① The plastic bottle is moved into the camera’s field of view.

② The plastic bottle rotates around its center, and the camera captures the image of the plastic bottle.

③ Visual processing feedback results to institutions.

Binarized image processing is performed on the plastic bottle identification area. By using defect detection and setting the upper and lower limits of the area, it can be judged whether there is surface scratches.

As shown in the figure on the right, the green mark in the blue identification box is the part.

Judged to be defective:

Remark:

① There is a bonding line on the surface of the plastic bottle, as shown in the figure below. Since the line is no different from the defect imaging effect, the identification area needs to avoid this line.

Processing Schematic

① There is a bonding line on the surface of the plastic bottle, and the identification area should be avoided. If a flaw appears on the line, there will be a situation where the flaw cannot be identified. If it is on this line, and the defect is parallel to this line, it will not be able to judge the defect.

②There are a lot of scratches or dust on the plastic bottle sent by the customer, and the area is similar to the limit area marked by the customer, which interferes with the visual inspection and has a risk of inspection. The incoming material must be clean and tidy, otherwise it can only be avoided by setting a large area identification value.

Black spot flaws are indistinguishable and blend in with the background. And the area is small. If this area is set as the upper and lower limits of recognition, it is easy to misjudge the surface dust as black spots, and there is a risk of detection.

As shown on the right:

Processing Schematic

Binarized image processing is performed on the plastic bottle identification area, and whether there is dust can be judged by using defect detection and setting the upper and lower limits of the defect area.

As shown in the figure on the right, the green mark in the blue identification box is the part.

Judged to be defective:

Remark:

① Dust on the bonding line and with a small area cannot be inspected.

②There are a lot of scratches or dust on the plastic bottles sent by the customer for testing, and the area is similar to the defect area marked by the customer, which interferes with the visual inspection and has a risk of inspection. The incoming material must be clean and tidy, otherwise it can only be avoided by setting a large area identification value.

When there are streaks on the plastic bottle, there will be black borders in some places. Through visual binarization, the black border can be distinguished from the background. If the black border is close to the background level, it cannot be extracted.

As shown in the figure, the thickness of the inside of the plastic bottle may be inconsistent (or affected by dust, fingerprints, etc.), and there is a black stripe background, which is highly overlapped with the stripe defect level, resulting in the risk of misjudgment in detection.

The extreme fringes sent by the customer for testing have shallow features and are not visible in the image.