Best Optimized Method

1. Make Image Tracking Optimized standard

1.1 Choose an appropriate picture size

The width-to-height ratio of the horizontal plate identification map should be controlled at about 3:2 or 4:3, and the vertical plate recognition image should be controlled at around 2:3 or 3:4; It is recommended that the size of the horizontal identification map is between 640x480 and 1920x1080, and the recommended size of the vertical recognition image is between 480x640 and 1080x1920;

1.2 Improve the recognition star rating of identification map

Please pay attention to the star feedback after the tracking image is uploaded to the platform. It is recommended to use images with 3 stars or higher;

1.3 A special way to improve recognition

Currently, WebAR only supports recognition and does not support tracking temporarily, so you can upload multiple pictures in the same scene to make it easier to recognize;

If we want to recognize the Pepsi bottle logo, we can add multiple types of identification map when creating a WebAR scene; for example:

Original standard 2D design draft

Photos taken in all directions

1.4 A special way to make physical identification

By taking pictures of the real object, and then uploading the photos after processing, the following is a real shot of multiple logo images of Beijing Hengtai Plaza, users can scan the physical logo at the entrance of Hengtai Plaza to complete the AR experience;

2. Improve the recognition star tutorial of the identification map

• Whether the image has sharp details. In the star rating of the identification map, a square contains 4 characteristic points, and a circle does not contain any characteristic points. When making an identification map, users try to avoid large areas of circular and sleek design.

Original image

Star rating result

• Properly increase the contrast of the image;

Original image

High-contrast image

Star Rating Result Updated image

• The image is only partially with sharp details and good high contrast. As shown in the figure below, it can be observed that the overall blueness of the left and upper parts does not have sharp details and good high contrast, but only has rich details in the lower right corner;

Original image

Star Rating Result

Uneven complete pictures and pictures with rich details after cropping (we cut out the parts with sharp details and good high contrast in the lower right corner of the picture separately for testing).

• We found that even if only the lower right corner image is retained, the star rating is still not optimistic. The reason is that although this part of the image contains enough sharp details and good contrast, there are repeated patterns that will hinder the recognition performance. , Lowered the star rating of the rating; in order to obtain the best effect, please choose an image without repeating patterns or an image without rotational symmetry;

Original image: repeated symmetrical patterns

Break repetitive symmetrical patterns

Star rating result: star comparison after breaking the repeated symmetrical pattern

• Try to avoid all kinds of fuzzy designs in large areas;

We will blur a large area around the original image in the image above

Star rating result: lowered by one star

3. Explanation of the reasons why it is sometimes difficult to identify or identify

• The content of the printed identification is inconsistent with the identification map uploaded in the background, and the problem of nonrecognition will occur, so please make sure that the content of the identification map uploaded in the editor and the identification map printed to scan for the same scene is consistent;

• Affected by the material of the printed paper during recognition, reflections, wrinkles, etc. may occur. These interference factors may cause recognition failure or difficulty. Therefore, please use diffuse and non-reflective materials to print the identification map to ensure that there are no reflective points, and Keep the test paper flat;

• If you use pictures taken by mobile phones, cameras and other equipment as the identification map, the identification map will be unclear due to the shooting angle, pixels, light and other factors, which will affect the recognition quality, so it is recommended to upload the original file or scan to ensure the greatest degree of assurance The clarity of the identification map;

• If the color tone of the uploaded identification map is relatively single, there are too many simple lines, or the pattern is not uniformly distributed, it may cause the phenomenon that it is not recognized or is difficult to recognize. Therefore, please choose a picture with rich color tone and more complex pattern as the identification map, and The overall distribution of the pattern should be as uniform as possible, which can effectively avoid the deviation of the recognition center of gravity and achieve a better recognition effect;

• A large number of small texts will affect the recognition effect of the image. If the uploaded identification map contains a large number of small texts, it may cause the phenomenon of not being recognized or difficult to recognize, so please choose a picture without a large number of small texts. Identification map;

• A large number of shadows will affect the recognition effect of the image. If the uploaded identification map contains a large number of shadows, it may cause the phenomenon that it is not recognized or difficult to recognize, so please choose a picture without a lot of shadows as the identification map;

• If the overall tone of the picture is dark, single, large areas of pure color, or even black and white tones, it may be difficult to recognize or difficult to recognize, so please choose a brighter picture as the identification map.