3
property arises from the circuitry of the event camera which
outputs the pixel intensity on a logarithmic scale. The high
temporal resolution arises from event cameras stream of
events instead of the whole frame. Due to these proper-
ties, event cameras are capable of responding to intensity
changes of the scene when the object is moving fast, which
means that event images almost do not suffer blur due to
fast motion. This type of vision sensor is aimed to be used
to overcome the issues in environments with poor lighting
and airborne dust as in underground mines. The principle
of operation of event cameras is based in dynamic scenes in
which the object of interest is moving, the camera is mov-
ing, or the lighting conditions is changing. Figure 3 shows
two types of images that have been captured by an event
camera. Because images from event cameras differ from
images from regular cameras, existing deep learning algo-
rithms cannot be implemented directly to event-camera
data driven tasks. In addition, there is no enough training
data for event images especially in underground mining
environments [27].
The main contributions of this research are:
• Building a deep learning-based binary image seman-
tic segmentation for underground roof bolting using
color images.
• Adapting the deep learning model to be able to
implement binary semantic segmentation for both
types of images images from regular cameras and
imaged from event cameras.
PROPOSED MODEL
The proposed model is based on formulating the auton-
omous roof bolting process as a binary model where the
goal is to classify the seen into two regions, rock regions or
non-rock regions. Among the three main tasks in computer
vision, image semantic segmentation is the most appropri-
ate method to classify rock areas from any non-rock areas.
Image segmentation using deep learning convolutional
neural networks model is promising to provide a high accu-
rate segmentation model. The authors selected to use a deep
learning convolutional neural network based image seg-
mentation model called Unet [28]. Unet is a convolutional
neural network-deep learning model which is used for
semantic segmented. The training data are images (color)
while the training labels are labeled masks. A labeled mask
is an image with a limited number of groups. Each groups
contains pixels of the same object. Figure 4 shows the pro-
posed deep learning based semantic segmentation model
using Unet which is composed of two parts, the training
part and the testing part.
Figure 2. A Training image with its corresponding ground-
truth training mask in Edgar Mine in Idaho Springs, CO.
This pair of images form the training image and the training
label for the proposed semantic segmentation model which
will produce a segmented image (image) for a given color
image
Figure 3. Two types of event camera images: (a) events image. (b) accumulated image
property arises from the circuitry of the event camera which
outputs the pixel intensity on a logarithmic scale. The high
temporal resolution arises from event cameras stream of
events instead of the whole frame. Due to these proper-
ties, event cameras are capable of responding to intensity
changes of the scene when the object is moving fast, which
means that event images almost do not suffer blur due to
fast motion. This type of vision sensor is aimed to be used
to overcome the issues in environments with poor lighting
and airborne dust as in underground mines. The principle
of operation of event cameras is based in dynamic scenes in
which the object of interest is moving, the camera is mov-
ing, or the lighting conditions is changing. Figure 3 shows
two types of images that have been captured by an event
camera. Because images from event cameras differ from
images from regular cameras, existing deep learning algo-
rithms cannot be implemented directly to event-camera
data driven tasks. In addition, there is no enough training
data for event images especially in underground mining
environments [27].
The main contributions of this research are:
• Building a deep learning-based binary image seman-
tic segmentation for underground roof bolting using
color images.
• Adapting the deep learning model to be able to
implement binary semantic segmentation for both
types of images images from regular cameras and
imaged from event cameras.
PROPOSED MODEL
The proposed model is based on formulating the auton-
omous roof bolting process as a binary model where the
goal is to classify the seen into two regions, rock regions or
non-rock regions. Among the three main tasks in computer
vision, image semantic segmentation is the most appropri-
ate method to classify rock areas from any non-rock areas.
Image segmentation using deep learning convolutional
neural networks model is promising to provide a high accu-
rate segmentation model. The authors selected to use a deep
learning convolutional neural network based image seg-
mentation model called Unet [28]. Unet is a convolutional
neural network-deep learning model which is used for
semantic segmented. The training data are images (color)
while the training labels are labeled masks. A labeled mask
is an image with a limited number of groups. Each groups
contains pixels of the same object. Figure 4 shows the pro-
posed deep learning based semantic segmentation model
using Unet which is composed of two parts, the training
part and the testing part.
Figure 2. A Training image with its corresponding ground-
truth training mask in Edgar Mine in Idaho Springs, CO.
This pair of images form the training image and the training
label for the proposed semantic segmentation model which
will produce a segmented image (image) for a given color
image
Figure 3. Two types of event camera images: (a) events image. (b) accumulated image