In DocuStudio, hotpoints can be automatically applied to a diagram using Optical Character Recognition ("OCR") algorithms that detect and apply hotpoints.
We support three standard auto-hotpointing algorithms:
- SVG Auto-Hotpointing: This algorithm reads in the SVG's XML, which should include x,y coordinates for where callouts exist on a diagram, and we use those coordinates to add hotpoints to the Documoto page. As long as the SVG contains 'good callout data', this algorithm auto-hotpoints with near 100% accuracy.
- Contour Shape Algorithm: This algorithm is an optical character recognition algorithm that looks for circles in the drawing, and then compares the numbers in those circles to the bill of material to determine if there are any item numbers that match the numbers in the callouts. If so, we add a hotpoint to the callout. This algorithm is dependent on the quality of the image and the callout font-readability, and auto-hotpoint accuracy varies depending on that. With high quality drawings, this algorithm can near 90% accuracy in many cases.
- Blob Finder Algorithm: This algorithm is an optical character recognition algorithm that looks in the drawing for any line that looks like an item number in the BOM, then adds hotpoints to the number. This algorithm is dependent on the quality of the image and the callout font-readability, and auto-hotpoint accuracy varies depending on that, but this is the least accurate in general due to the fact that lines often could look like numbers (ex. 1's typically do hotpoint with less accuracy).
Note: auto-hotpointing is not 100% accurate when hotpointing
raster images (.jpg, .png, .gif, etc).
The OCR algorithms rely on the quality of the callout numbers in the image to map callouts numbers to item numbers in the bill of materials. For example, the algorithm(s) may apply numbers like '7' or '1' to lines in the diagram that look like they could be item numbers.
Note: unless using an .svg image, there is always some manual
review and repair required for hotpoints.
Comments
0 comments
Please sign in to leave a comment.