An automatic warehouse inventory monitoring system based on text extraction using computer vision

Nowadays, most of product companies around the world do sales forecasting and inventory management to give themselves benefits to be able to take care of all customers’ demand and avoid the shortage of supplies that might cause a lost in sales. Most companies then require at least a warehouse whether it is an in-house or outsourced warehouse to store products safely and be able to deliver them to assigned destinations. That is why warehouse management is an inevitable factor when we consider maximizing supply chain’s profit. In order to ensure that operating warehouses perform product storage correctly, companies usually request for an inventory checking report from warehouses, some for annually and some for semi-annually. This report normally presents meaningful keys that represent identity of each product such as product ID, pallet’s tag ID, location of pallet on the rack, and quantity of products in that pallet. However, this request indirectly sends non-benefit impacts to the warehouse since inventory checking has been done manually which requires several workers to unload, check, and load pallets back on the rack and also needs to stop all transporting schedule to do the checking. With the following impacts, warehouse needs to reschedule its normal activities and sometimes even postpone the delivery appointment. This study is going to find the solution for these drawbacks by replacing manual inventory counting with an automatic warehouse inventory monitoring system based on text extraction with the use of computer vision that can detect product label attached on every pallet on the rack, recognize and extract key information from label, and send digital text to store in document for comparing with the master data. Under the field of computer vision, technology of image processing will be applied in the part of detecting product label from captured images while optical character recognition will be applied to extract text from cropped image of label. With the help of technology, warehouse can reduce time used and decrease human error in the inventory checking process.