Buzzard is a Deep Learning solution for real-time defect detection. The innovative board scanning solution was named after a common hawk species that has eight times better vision than human. Buzzard is a research and engineering success that was created by our own product development in order to even better serve our customers.
In 2017, with recent advances in Machine Learning and Deep Leaning, an idea of enhancing our board scanning system with the power of autonomous artificial agents was born. The idea then grew into a project known as Buzzard and resulted in ground-breaking product development in sawn timber grading.
Detecting rotten area on timber surface has known to be extremely challenging due to its high variations in color, shape and textural appearance. In order to separate rotten area from reaction wood, sunburn and other color defects, a more effective feature extraction technique was in high demand. The difficulty in rot detection inspired and motivated the development of Buzzard which has later proven to be extremely powerful in solving the problem.
Buzzard is in general based on massive amount of data consisting of images of boards and defects on them as well as information on defect locations on the boards. In early stages, a huge team effort was put into analyzing our customers’ requirements and constructing an efficient data pipeline that enabled further development of the entire solution. Buzzard was initially created by our Data Scientist Glen Guo, who later on was joined by our software engineer Nikolay Rudakov in order to further develop the solution and make Buzzard even more powerful.
Buzzard is currently used to detect various defects on timber surface. Buzzard detects defects like rot and a more specific defect bird eye that is caused by needles grown into the wood material in Radiata Pine wood species in South America. For our customers, Buzzard comes as a standard functionality in our BoardMasterNOVA scanners and provides significantly better defect detection accuracy without compromising on production capacity.
The first version of Buzzard was released in October 2017 when it tested and then used for the first time in production at NK Lundströms sawmill in Sweden. The tests were a success and their impressive results encouraged and motivated further investments in AI in order to accelerate Buzzard’s further development. The first sawmill to have Buzzard in production use for bird eye detection was Maderas Arauco Nueva Aldea sawmill in Chile. BoardMasterNOVA grading system with the Buzzard solution was installed and commissioned at the sawmill’s dry sorting line in the beginning of 2019 (read more about the birde eye detection in Nueva Aldea).
Nowadays, approximately 29 BoardMasterNOVA systems in six different countries are using the latest Buzzard solution in defect detection. Buzzard is continuously being further developed in order to exceed our customers’ needs and requirements. Our product development is constantly adopting cutting-edge technologies in our products so that our scanning systems will provide maximum benefits to our customers.
FinScan’s customer-oriented product development, combined with a solid knowledge of the latest technologies has resulted in modern and efficient solutions for the sawmill industry over the years. Now our latest product development for BoardMaster, the bird eye detection implemented with neural networks, is the first in the world in production use.
The bird eye detection was developed for market specific needs to Maderas Arauco Nueva Aldea sawmill in Chile. Bird eye defects are a crucial defect type in Chile’s main wood species radiata pine. FinScan has already utilized deep neural networks in their systems and now the technology is used for bird eye detection. The bird eye detection uses deep neural networks to detect bird eye from non-planed and planed wood. The bird eye detection ensures high grading accuracy, improved defect detection and delivers more reliable optimization results. The time-consuming defect detection training has already been done by our product development. The system doesn’t require any detection training from customer’s side and it can be taken into use directly.
The BoardMasterNOVA was delivered to Maderas Arauco Nueva Aldea sawmill in Chile in 2018. It was FinScan’s second delivery to South-America. The first FinScan grading system was delivered to Uruguay in 2016. Contract with Maderas Arauco was made in summer 2018 and delivery was made in the end of 2018. BoardMaster was installed and taken into use in the sawmill’s dry sorting line in the beginning of 2019.
Maderas Arauco is the most significant producer of sawn timber in Chile and the eighth biggest producer of sawn timber in the world. Maderas Arauco produces approximately 4,5 million m³ of lumber per year. The company has eight sawmills in Chile and one in Argentina. Maderas Arauco is constantly updating its production lines. Increased speed and human errors in production were main reasons for the company to invest in automated grading system in their sawmill. FinScans BoardMaster was chosen from possible solutions due to its excellent grading results.
Canadian Forest Industries shares the newest scanner and optimizer technology on the market: https://www.woodbusiness.ca/a-look-at-the-latest-scanners-and-optimizers
BoardMasterNOVA is a non-turning auto-grading scanner for sawmills and planer mills. BoardMasterNOVA can provide a complete analysis of green and dry boards, performed in 10 different directions including accurate defect detection and reliable optimization results. It supports WWPA, NLGA and ALS grading rules for dimensional lumber, common grades, clear cut components and shop grade. BoardMasterNOVA has comprehensive graphical reporting tools for operators, production planners and top management. FinScan’s auto-grading systems can be easily installed in the existing or new grading lines.