Measuring Seed Velocity and Seed Counting in a Pneumatic Conveying System using Electrostatic Sensors
Date
2017-09-26
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
0000-0002-4356-0783
Type
Thesis
Degree Level
Masters
Abstract
The trend towards precision agriculture has led to the advent of a new generation of
modern equipment for agriculture. Pneumatic conveying systems are widely used for seeding
operations in modern agriculture. This thesis considers the problem of seed velocity
measurement and counting seed number in the pneumatic conveying process.
For the
owing seed velocity measurement, the experimental object was wheat seed.
Tests were performed with air velocities of 12, 15, 20, 25, and 30 m/s in a 57.3 mm acrylic
pipe while the seed mass
ow rate was increased from 1 kg/min to 6 kg/min in 1 kg/min
increments. All measurements were taken 10 meters downstream of the feed point of the
rotary feeder into the air steam. The proposed method of velocity measurement is based on
the cross-correlation algorithm. Two di erent active start points of the cross-correlation have
been developed, one is the xed time window, and the other is the threshold detection. The
horizontal velocity of the seeds and the slip ratio were calculated from the results. Beside
some clumping-seed testing groups, the slip ratio between the total seeds velocity and the
air velocity was relatively constant at approximately 0.63.
For counting the number of seeds, the experimental objects were wheat and canola, and
the tests were taken in the secondary pipeline of the seed drill system. One contribution
of this part is signal denoising using compressive sensing. Compressive sensing provides a
feasible method based on the sparsity of the seed signal. The other contribution of this part
is utilizing pattern recognition technique for counting number. Features were extracted from
the seed signal, which are Threshold Detection, Full Width Half Maximum, Cluster Width,
Peak and Valley Detection, Number of Turns, and Energy Comparison. Four of them was
selected to be applied in the experiments. Multiple classifi er approach was also developed in
the classifi cation task of the pattern recognition. The counting accuracy for all of the wheat
groups were higher than 92% and for the canola groups were higher than 96%.
Description
Keywords
Velocity Measurement, Counting, Electrostatic Sensor, Signal Denoising, Pattern Recognition, Cross-Correlation
Citation
Degree
Master of Science (M.Sc.)
Department
Electrical and Computer Engineering
Program
Electrical Engineering