Parcel tracking solution

One of the leaders in the US logistics market. The company's main focus is on smart solutions that optimize operating processes and improve customer satisfaction rates.

Project background  

The rise of e-commerce, especially cross-border e-commerce, has dramatically changed the climate of the postal and parcel industry. Operators are focusing on customer engagement at every touchpoint of their journey. 

Most postal and parcel delivery companies have started to invest more in technological solutions in order to provide connected logistics. Some of the solutions include: 

  • real-time tracking,  
  • automation of distribution,  
  • smart shipment and asset management,  
  • and better data management through the use of visualization tools.  

The main idea of the project was digitalization and automation of manual labor, in order to optimize resources, as well as the speed of processing postal information. The initial solution involved much manual human work, which cost too much and was time consuming.  

Bar codes on parcels often get mechanical scratches or other damages before reaching the sorting point.  

Such defects obstruct the correct identification of a parcel in the system, causing delivery mistakes. The client needed to prevent possible mistakes caused by damaged bar codes and also reduce human-made mistakes. 

Duration
  • June 2019 – October 2019
Location
  • The USA
Industry
  • Logistics
Development team
  • 2 С++ Developers
  • 2 Python Developers
  • 1 PM
  • 1 Designer
Technologies
  • Python
  • Optical Character Recognition (OCR) with Tesseract
  • С++
  • NumPy
  • OpenCV
  • scikit-image
  • TensorFlow
  • Keras
  • RNN
  • Darknet
  • Yolo
  • CNN-RNN-CTC
  1. Correct detection of the postal label among many others
  2. Extraction of country and zip codes and text if the bar code is not readable
  3. Compositing the delivery information
  4. Checking if the postage information is immediately transferred into the database
  5. Creating instructions for automatic sending of the parcel to the proper vehicle for further delivery
Business needs
  1. To receive professional consultation on how to improve the current business processes
  2. To validate the idea and prove the suggested automation concept
  3. To improve the detection of parcels under conditions of poor light at the premises and the state of damaged parcels
  4. To detect the label and recognize the bar code and various fields on the parcel label
  5. To increase the percentage of correctly recognized addresses and reduce the percentage of errors in sending
  6. To get the opportunity to scale the solution

Solution

CHI Software team started work on this project validating the client’s idea and proving the concept. The main tasks were detection and recognition of objects and image preprocessing. We also selected and trained neural network models to ensure that parcels, letters, and other items of interest are identified automatically and sorted correctly.  

We were responsible for the backend part of the project. Our design team was only involved in minor tasks connected with the UI for parcel station operators.   

Tech block

Tech stack: Python, Optical Character Recognition (OCR) with Tesseract, С++, NumPy, Pillow, OpenCV, scikit-image, TensorFlow, Keras, CNN, RNN, Darknet, Yolo, CNN-RNN-CTC 

Tesseract  an optical character recognition (OCR) tool in Python. We use Tesseract to detect embedded characters in an image. 
Pillow  a Python Imaging Library (PIL), which adds support for opening, manipulating, and saving images. The current version identifies and reads a large number of formats. 
TensorFlow  artificial intelligence framework for neural networks development. It allows the creation of large-scale neural networks with many layers. TensorFlow is used for: Classification, Perception, Understanding, Discovering, Prediction and Creation 
Keras  used for creating deep models which can be productized on smartphones. Keras is also used for distributed training of deep learning models. 
Yolo   You only look once (YOLO) is a state-of-the-art, real-time object detection system. 
Darknet   an open source neural network framework written in C and CUDA. It is used to classify images. 
Client values
  1. We proved that it is possible to automate the current process with easy-to-use, powerful automation of the business processes
  2. We optimized work of distribution centers worldwide
  3. We ensured that parcels, letters, and other items are identified automatically and sorted properly
  4. We demonstrated how to scale the current solution
  5. CHI team’s solution also helped the client reduce the overall business costs

Testimonials

quote

Working as a Project Manager, I like challenging tasks and good communication. Parcel Station is surely among my favourite projects in terms of complexity and delivered results. Our inhouse team and AI/ML expertise allows us to cope with automation projects rather fast. We were able to validate the client’s idea of improving items detection and recognition and advise on further steps. CHI Software team trained the custom neural network to correctly identify bar codes on parcels even if any mechanical damages have occurred. Glad that our solution helped the client reduce the overall business costs and reach their business goals.

quote
Elena Morozova Project Manager
quote

Let’s bring your idea to
life together!

    img img Successfully applied!