Vision-based Weed vs Crop Discrimination for Selective Weeding The - - PowerPoint PPT Presentation

▶

vision based weed vs crop discrimination for selective

Vision-based Weed vs Crop Discrimination for Selective Weeding The - - PowerPoint PPT Presentation

Jan 19, 2023 151 likes •570 views

Vision-based Weed vs Crop Discrimination for Selective Weeding The UK Onion & Carrot Conference & Trade Exhibition Petra Bosilj University of Lincoln, UK L-CAS Notthingham, November 14 th 2017 Notthingham, November 14 th 2017 Petra

slide-1

SLIDE 1

Vision-based Weed vs Crop Discrimination for Selective Weeding

The UK Onion & Carrot Conference & Trade Exhibition

Petra Bosilj

University of Lincoln, UK L-CAS

Notthingham, November 14th 2017 Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 1 / 12

slide-2

SLIDE 2

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-3

SLIDE 3

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-4

SLIDE 4

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-5

SLIDE 5

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-6

SLIDE 6

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-7

SLIDE 7

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-8

SLIDE 8

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-9

SLIDE 9

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-10

SLIDE 10

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-11

SLIDE 11

Motivation

project by: University of Lincoln, Garford Farm Machinery Ltd. (sponsored by Innovate UK and BBSRC) locate and recognize type of vegetation in the image segmentation (soil removal), classification (identifying the plant type)

selective treatment (spraying) of weeds vision-guided robotic hoeing (mechanical)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 2 / 12

slide-12

SLIDE 12

Outline

1

Motivation and goals

2

Our approach Data collection and preparation Segmentation Classification

3

Conclusions

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 3 / 12

slide-13

SLIDE 13

Our approach

Image Processing (attribute morphology):

analysing and selecting contrasted objects in the image based on their characteristics: shape, texture, color, neighbourhood local processing, robust to different lighting

Machine Learning techniques:

determining the type of plant

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 4 / 12

slide-14

SLIDE 14

Data collection

two-camera system: near infra-red (NIR) + color

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 5 / 12

slide-15

SLIDE 15

Data collection

two-camera system: near infra-red (NIR) + color mounted on a manually operated setup

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 5 / 12

slide-16

SLIDE 16

Data collection

two-camera system: near infra-red (NIR) + color mounted on a manually operated setup (for now)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 5 / 12

slide-17

SLIDE 17

Data collection

two-camera system: near infra-red (NIR) + color mounted on a manually operated setup (for now)

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 5 / 12

slide-18

SLIDE 18

Data collection - examples

examples of carrot and onion fields in different stages of growth

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-19

SLIDE 19

Data collection - examples

examples of carrot and onion fields in different stages of growth

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-20

SLIDE 20

Data collection - examples

examples of carrot and onion fields in different stages of growth

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-21

SLIDE 21

Data collection - examples

examples of carrot and onion fields in different stages of growth carrots

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-22

SLIDE 22

Data collection - examples

examples of carrot and onion fields in different stages of growth carrots

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-23

SLIDE 23

Data collection - examples

examples of carrot and onion fields in different stages of growth carrots

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-24

SLIDE 24

Data collection - examples

examples of carrot and onion fields in different stages of growth carrots

nions

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-25

SLIDE 25

Data collection - examples

examples of carrot and onion fields in different stages of growth carrots

nions

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 6 / 12

slide-26

SLIDE 26

Data preparation

NIR + color image registration: image alignment

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 7 / 12

slide-27

SLIDE 27

Data preparation

NIR + color image registration: image alignment

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 7 / 12

slide-28

SLIDE 28

Data preparation

NIR + color image registration: image alignment calculating NDVI (normalized difference vegetation index) image

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 7 / 12

slide-29

SLIDE 29

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions

clean patch of onions

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-30

SLIDE 30

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions

nions with weeds

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-31

SLIDE 31

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions

GROUND TRUTH OTSU RATS MAX-TREE

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-32

SLIDE 32

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions

GROUND TRUTH OTSU RATS MAX-TREE

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-33

SLIDE 33

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions good performance on low-content images

artificial low vegetation example

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-34

SLIDE 34

Vegetation segmentation

typical: threshold determined globally based on image histogram, gradients

poor performance on textured background, noise removal required

proposed approach: locally selected regions good performance on low-content images

GROUND TRUTH OTSU RATS MAX-TREE

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 8 / 12

slide-35

SLIDE 35

Grouping into objects

The segmentation output is organized into a list of separate objects

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 9 / 12

slide-36

SLIDE 36

Grouping into objects

The segmentation output is organized into a list of separate objects Objects typically correspond to one plant or a few overlapping plants

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 9 / 12

slide-37

SLIDE 37

Classification

Object-based classifier applied to the list of regions Regions classified as crop, weed and mixed (for overlapping objects) Distinguishing features: shape (circularity, elongation), texture (color variance)

GROUND TRUTH CLASSIFICATION OUTPUT

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 10 / 12

slide-38

SLIDE 38

Classification

Object-based classifier applied to the list of regions Regions classified as crop, weed and mixed (for overlapping objects) Distinguishing features: shape (circularity, elongation), texture (color variance) SVM (Support Vector Machines) and RF (random forest)

GROUND TRUTH CLASSIFICATION OUTPUT

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 10 / 12

slide-39

SLIDE 39

Classification

Object-based classifier applied to the list of regions Regions classified as crop, weed and mixed (for overlapping objects) Distinguishing features: shape (circularity, elongation), texture (color variance) SVM (Support Vector Machines) and RF (random forest) Overlap processed later, using pixel-based approaches

GROUND TRUTH CLASSIFICATION OUTPUT

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 10 / 12

slide-40

SLIDE 40

Conclusions

Plant detection and recognition system that works under field conditions Adaptable to different lighting conditions, amount of vegetation content, different crop types Local decisions based on analysis of region characteristics

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 11 / 12

slide-41

SLIDE 41

Thank you for your attention!

Questions?

Petra Bosilj (UoL, L-CAS) Notthingham, November 14th 2017 12 / 12