Jet Infosystems together with Segezha Group created service for measurement of volume of the prepared wood

2020: Creation of service for measurement of volume of the prepared wood

On July 20, 2020 the Jet Infosystems company reported that it jointly with specialists of Segezha Group created service for measurement of volume of the prepared wood. This system developed using video analytics and machine learning (ML) reduces time of assessment of volume and quality of raw materials and does process of logging by more controlled.

Jet Infosystems together with Segezha Group created service for measurement of volume of the prepared wood

According to the company, during procurement of the wood it is exported by lorries (timber carrying vessels) on warehouses to deliver to plants. One of key tasks in this process – to make the objective analysis of the arriving raw materials and to make exact measurement of volume and quality of logs. The Segezha PPM Segezha Group (enters into AFK "Sistema") in the Republic of Karelia decided to carry out a pilot project to solve this problem.

Specialists of Jet Infosystems IT company created model which allows to measure precisely the dense volume of the timber arriving on point of acceptance on timber carrying vessels. A system is based on technologies of computer vision (video analytics and machine learning), works based on deep neural networks for the analysis of images and the sequences (Convolutional and Recurrent Neural Networks).

Jet Infosystems together with Segezha Group created service for measurement of volume of the prepared wood

The principle of its work is that the loaded timber carrying vessel undergoes scanning on a photoframe of check-point (scan track) where cameras take a set of pictures; at the same time models of machine learning analyze a load, define breed and other characteristics of a tree, and consider wood volume before its transportation on paper plant. On the monitor the operator sees specific packs with indication of the data (breed, quality, diameter) received in the automated mode.

Thus, a system solves problems of decrease in influence of a human factor on expert evaluation, reduces time of assessment and, as a result, the general logistic process. A system helps to create objective metrics of quality of raw materials for the further analysis.

Check of work of methods of computer vision for detection of logs and ML in integration with the available information systems are important steps on the way to industrial application of scalable services. Pavel Vakhnin, the board member, the vice president for information technologies and process automation of Segezha Group noted

The system developed by a command Jet Infosystems adapts under any weather conditions (snow, a rain, dirt, a bright sun), will recognize incorrect laying of packs (if distance in accordance with GOST less than 0.3-0.5 meters), has the good performance of processing of pictures and automatically determines illegible photos.

For July, 2020 the accuracy of comparison of photos and wood packs is recorded at the level of 99% (it is checked on nearly three thousand timber carrying vessels), and the accuracy of determination of breed of wood — 99%. And data of one timber carrying vessel passing through the scan track are processed within several seconds whereas earlier process could borrow up to X minutes with more low accuracy.

Results of a pilot project suggest that use of modern technologies are possible even in such areas as the visual analysis of the arriving raw materials. Evgeny Kolesnikov, the head of the Center of machine learning told Jet Infosystems

The developed service creates the general standard of application ML in projects and can be used for other platforms, types of raw materials and methods of delivery.