The high-pass identification map is uploaded to the background, we can see the star rating, the star value of the maximum of 5 stars, the more the star value of the following benefits:
1, the identification of the more easily identify the map
2, the identified objects will not appear jitter
3. Faster recognition speed
And sometimes we upload only 2 stars, or less, and this time we need to deal with recognition to achieve the effect we want,
From the above 5-Star recognition image we can summarize:
1. The details of the identification chart are angular and angular, and the number of edges is extra.
2, the edges and corners of the identification chart are evenly divided
3, the picture single element is very small
The labeling of the characteristic points of a high-pass identification chart has the following rules:
1, round the entity no characteristics point
2, the square of the entity has four characteristics of the point
3, semi-circular entities have two characteristic points
Let's take a look at the contrast picture effect:
As can be seen from Figure 1, the first picture has a large circle of elements, the corresponding recognition characteristics are very few,
The second picture square, or more sharp line, image recognition characteristics of a lot of points, and evenly distributed
And sometimes, we upload pictures clearly the image distribution is very complex, but still up to 5 stars, in addition to composition elements, there are
An important parameter that has a big relationship with the clarity of the local image:
The clearer the local picture, the easier it is to find the recognition feature point
If we upload the picture ready-made, can not be changed, through the following comparison chart to solve:
The first picture in Figure 4, because it is too vague to identify only 1 stars,
The second picture, we changed the background of the picture, the recognition intensity increased to 2 stars,
In the third picture, we increased the image quality to 3 stars.
In Figure 5 We continue to enhance the image quality,
The first picture, enhanced the contrast of the picture, the recognition coefficient of 4 stars
The second picture, we removed the background color of the original picture, this is another solution
Figure 6 The image recognition coefficient is all 5 stars, the picture has been optimized to become clear, angular
Other additional identification chart making instructions:
1. Avoid uneven distribution of picture recognition points (Figure 7)
2. Avoid identifying the elements in the graph are very round (Figure 8)
3. Avoid very regular pattern graphics (Fig. 9)
All right, here's how the Qualcomm identification map is handled.
Cross-platform AR augmented Reality Development (IV) (Qualcomm identification Figure 5 Star processing mode)