Warehouse Boxes Inventory Count
My project aim is to be able to count the number of boxes within a image. These boxes will always have a label pasted on them consisting of relevant barcodes which need to be read using computer vision.
Do you have any suggestions on what could be done to do an accurate count of the number of boxes?
One constraint is that these boxes can be of many different sizes and hence if I were to use hough transform, I must be able to cater for these different shapes.
Should I focus on counting the number of labels which will tell me the number of boxes, or should I use the contours of the boxes as the feature to count the number of boxes.
Could someone guide me in what algorithms I start playing with to achieve inventory count?
Up until now, I have tried the watershed algorithm.
One example image would be this
http://www.california-air-cargo.com/b...
Thank you!
what about the boxes that are in the 2nd and 3rd rows which are not visible in the Front view? What about the background & color of the box? is that constant?
If your real data is similar to that one, go for an edge distrance transform. Use peaks as seeds for watershed segmentation, then use results of both side views to calculate the total number of boxes.
Or take a look at this rubiks cube analyzer!