Performance: YUYV, clear camera image quality (high recognition accuracy), high performance, and resource usage.
Stability: MJPG, average camera image quality, low performance, and resource usage.
Camera Angle: It is recommended to fix the camera at an angle (avoid vertical positioning).
Remove Protective Film: Remove the camera's protective film and cover (if the recognition area is blurry or black) to ensure clear imaging.
Cropping Tips: The four-point positioning of the recognition area can slightly exceed the scale pan to capture the entire product.
No Movement After Cropping: Do not adjust the camera's recognition area position during use, as it will directly affect the accuracy of product recognition.
Ensure the correct camera is selected in the options (the software defaults to the first camera).
Check if the camera is offline.
Verify the camera is properly connected at the physical layer (check if the device appears in the device manager).
Ensure the camera is not being used by other software (test with a camera tool).
Check for unstable camera connections.
Remove Film: If the scale pan has an opaque protective film, remove it before using the device. Learning with the film on (and later removing it) or learning without the film and then removing it can lead to poor recognition results and inconsistent data across devices.
Color: Optimal choices: White > Silver > Blue > Black. White matte or metallic colors are recommended; avoid black or reflective materials.
Texture: Optimal choices: Matte > Glass (reflective materials).
Optimal Conditions: Normal lighting > Shadows > Overexposure (spotlights/colored lights).
Maintain normal lighting; shadows and overexposure can prevent the camera from correctly capturing product colors and textures, reducing recognition accuracy.
Place the scale in a well-lit area but avoid overexposure.
Product Positioning: Place the product in the center of the scale pan to ensure the camera captures the entire product.
Opaque Bags: Open the bag (facing the camera) to expose the product's features for recognition.
Bagged Learning: During learning, use bagged products to improve recognition accuracy.
Basket Learning: If products are recognized in a basket, it is recommended to include the basket during learning.
Antivirus software cleaning local folders.
Inconsistent product data transmission protocols (changes in product name/PLU).
Learning data migration: inconsistencies in product names, PLU numbers, and hardware devices.
Inability to obtain weight/weight not meeting recognition trigger conditions.
