Weed Detection in Crops Using Computer Vision Presenter: Dr. Yasir - - PowerPoint PPT Presentation

▶
weed detection in crops
SMART_READER_LITE
LIVE PREVIEW

Weed Detection in Crops Using Computer Vision Presenter: Dr. Yasir - - PowerPoint PPT Presentation

Feb 16, 2024 19 likes •709 views

Centre of Robotics Weed Detection in Crops Using Computer Vision Presenter: Dr. Yasir Niaz Khan Researchers: Taskeen Ashraf, Danish Gondal, Novaira Noor. http://cs.ucp.edu.pk/index.php/robotics-security/ UCP Robotics Group Centre of Robotics

slide-1
SLIDE 1

Centre of Robotics

Weed Detection in Crops Using Computer Vision

Presenter: Dr. Yasir Niaz Khan Researchers: Taskeen Ashraf, Danish Gondal, Novaira Noor. http://cs.ucp.edu.pk/index.php/robotics-security/

slide-2
SLIDE 2

Centre of Robotics

UCP Robotics Group

  • Faculty
  • Dr. Yasir Niaz Khan
  • Dr. Syed Atif Mehdi
  • Dr. Musharraf Hanif
  • Dr. Oumeir Naseer
  • Muhammad Awais
  • Researchers
  • Aamir Ishaq
  • Sibtain Abbas
  • Ruhan Asghar
  • Hamad ul Qudous
  • Noman Saleem
  • More than 50

undergrad students

slide-3
SLIDE 3

Centre of Robotics

Road Map

  • Introduction
  • Problem Statement
  • Methodology
  • Experimentation & Results
  • Comparison
  • Conclusion
  • Future Work
slide-4
SLIDE 4

Centre of Robotics

INTRODUCTION

slide-5
SLIDE 5

Centre of Robotics

Importance of Rice Crop1,2

  • Feeds over 50% of World’s population
  • Pakistan 13th world wide in Rice production
  • Pakistan 4th in Rice Exports
  • Stands second in terms of staple food in Pakistan
  • 13% to the total value of Exports
  • Stands third in terms of cultivation area3

1. Old.parc.gov.pk, "NARC-Rice||Introduction", 2015. [Online]. Available: http://old.parc.gov.pk/NARC/RiceProg/Pages/intro.html. [Accessed: 20- Dec- 2015]. 2. Bayercropscience.com.pk,. 'Bayer Cropscience - Pakistan : Rice'. N.p., 2015. Web. 14 May 2015. 3. Fao.org,. 'Fertilizer Use By Crop In Pakistan'. N.p., 2015. Web. 18 June 2015

slide-6
SLIDE 6

Centre of Robotics

What are Weeds and Weed Control?1

Weeding Approaches Manual Hand weeding Hand hoeing Partially automated Herbicides application Biological means

1. Pakissan.com,. 'Integrated Weed Management In Rice :: Pakistan Agricultural News Chennal-:PAKISSAN.Com:- '. N.p., 2015. Web. 17 May 2015. 2. Eap.mcgill.ca,. "Biological Control Of Weeds". N.p., 2015. Web. 31 sep. 2015.

Biological Means2 1. biological spray

  • spore

suspension of an endemic fungus 2. a fish, the white amur

slide-7
SLIDE 7

Centre of Robotics

Issues with Weed Control Methods

  • Difficult to harvest
  • Disadvantage of uniform spraying

▫Uneconomical ▫Affects crop health ▫Environmental Pollution ▫Resistance to sprays

slide-8
SLIDE 8

Centre of Robotics

Resistance to sprays

Survey website at http://www.weedscience.org on September 13th, 2015.

slide-9
SLIDE 9

Centre of Robotics

Weeds are a Problem

  • Weed destroys 15-20% or in some

cases up to 50% of the crop1

  • Uniform spraying is uneconomical
  • Control Period of weeds is first 40-50

days

  • 1. Pakissan.com,. 'Integrated Weed Management In Rice :: Pakistan Agricultural News

Chennal-:PAKISSAN.Com:-'. N.p., 2015. Web. 17 May 2015.

slide-10
SLIDE 10

Centre of Robotics

Problem Statement

slide-11
SLIDE 11

Centre of Robotics

Problem Statement

“Automated localized weed detection in rice fields to avoid excessive uniform spraying; that will result in high, good quality yield with low production cost.”

slide-12
SLIDE 12

Centre of Robotics

Problem Statement

“Automated localized weed detection in rice fields to that will result in high, good quality yield with low production cost.”

slide-13
SLIDE 13

Centre of Robotics

Problem Statement

“ in rice fields to that will result in high, good quality yield with low production cost.”

slide-14
SLIDE 14

Centre of Robotics

slide-15
SLIDE 15

Centre of Robotics

Initial Experimentation

  • Testing conducted with few images

(DS-1)(1,2)

  • Broadleaf and sedges

1. Jircas.affrc.go.jp,. 'JIRCAS cyperus Difformis plants In Lowland Savanna Of West Africa'. N.p., 2015.

  • Web. 10 May 2015.

2. Mikobi.deviantart.com,. 'Water Lily In The Rice Paddies Around Angkor Wat'. N.p., 2015. Web. 10 May 2015.

slide-16
SLIDE 16

Centre of Robotics

Initial Experimentation

  • Three techniques

▫Based on localized FFT and Edge Detection1 ▫Based on localized Entropy ▫Based on Wavelet Transform2

10 20 30 40 50 60 70 80 90 100 Localized FFT Localized Entropy Discrete Wavelet Transform

Comparison Using Accuracy and FPR

Accuracy FPR

Accuracy: 89.60 %

Techniques Accuracy FPR FFT 74.85% 27.83% Entropy 76.66% 24.09% Wavelet 89.60% 17.50%

  • 1. Nejati, Hossein, Zohreh Azimifar, and Mohsen Zamani. "Using fast fourier transform for weed detection in corn fields."

Systems, Man and Cybernetics, 2008. SMC 2008. IEEE International Conference on. IEEE, 2008

  • 2. Noor, Novaira, and Yasir Niaz Khan. 'Weed Detection In Wheat Fields Using Computer Vision'. Graduate. FAST-NU

Lahore, 2014. Print.

slide-17
SLIDE 17

Centre of Robotics

Experimental setup & Dataset

slide-18
SLIDE 18

Centre of Robotics

Experimental Setup

  • Setup

▫MATLAB 2014 64 bit ▫Windows 8 64 bit ▫4 GB RAM ▫Core i5 1.70GHz Processor ▫LibSVM and RF

  • Dataset

▫Images taken height of 2-4 ft. ▫Angle of capture is 90 degrees ▫Image resolution is 1920x1080

slide-19
SLIDE 19

Centre of Robotics

Technique 1

Using Wavelet Transform involving Blur Detection

slide-20
SLIDE 20

Centre of Robotics

Overall Technique

Video Extract Every Nth Frame(Image) Blur Detection Module Weed Detection Module Output image Calculate Weed Coverage Trained SVM Model

Blur Non-Blur

slide-21
SLIDE 21

Centre of Robotics

Blur Detection

Dataset blur/Non-Blur labelled images Get image one by one Convert RGB to Gray Calculate Discrete Laplacian Extract Features Train SVM (Batch Training) Linear SVM Model

Min, max, std

slide-22
SLIDE 22

Centre of Robotics

Weed Detection

Input image Excessive green image Wavelet Transform Thresholding on Diagonal Coefficients Inverse Wavelet Transform Dilation Remove small regions Output image

slide-23
SLIDE 23

Centre of Robotics

Steps 1-3

Original Image Excessive Green Image Diagonal Coefficient Diagonal Coefficient(Filtered)

slide-24
SLIDE 24

Centre of Robotics

Steps 4-5

Dilation

slide-25
SLIDE 25

Centre of Robotics

Accuracy & FPR

  • Total Frames = 1717
  • Total Frames processed = 172
  • Non-blur frames detected = 67
  • Accuracy of blur detection = 84.88%
  • FPR of blur detection = 18.46%
  • Weed Detection Accuracy = 68.95%
  • FPR = 12.69%
  • Weed Detection Accuracy after blur removal =

76.16% (8% increase)

  • FPR = 13.38%
slide-26
SLIDE 26

Centre of Robotics

Weakness

Accuracy drops drastically when texture difference decreases with the growth of grass

slide-27
SLIDE 27

Centre of Robotics

Technique 2

Using SVM and Random forest with Moments

slide-28
SLIDE 28

Centre of Robotics

Dataset-2 Density Based

slide-29
SLIDE 29

Centre of Robotics

Weed Detection

Density Based Dataset Extract Green channel from RGB Calculate Mean,variance,kurtosis,skew Train Classifier (Batch Training) Calculate n-fold cross validation

Calculate complex moments

slide-30
SLIDE 30

Centre of Robotics

Accuracy Using First Four Moments

68.00 70.00 72.00 74.00 76.00 78.00 80.00 82.00 84.00 86.00 88.00 1 2 3 4 5 Accuracy

  • No. of Iterations

Linear kernel RBF kernel Random Forest

Accuracy: 82.22% RBF Kernel SVM C=8, g=0.25

slide-31
SLIDE 31

Centre of Robotics

Accuracy Using Complex Moments

Accuracy: 81.42% random forests with 300 trees

66.00 68.00 70.00 72.00 74.00 76.00 78.00 80.00 82.00 84.00 86.00 1 2 3 4 5 Accuracy

  • No. of Iterations

Linear kernel RBF kernel Random Forest

slide-32
SLIDE 32

Centre of Robotics

Accuracy Using Combined Moments

Accuracy: 86.06% RF With 300 trees

70 72 74 76 78 80 82 84 86 88 90 1 2 3 4 5 Accuracy

  • No. of Iterations

Linear kernel RBF kernel Random Forest

slide-33
SLIDE 33

Centre of Robotics

Comparisons

Accuracy and Execution Time

slide-34
SLIDE 34

Centre of Robotics

Accuracy

10 20 30 40 50 60 70 80 90 100 Linear kernel RBF kernel Random Forest Accuracy Type of classifiers Moments Feature set GLCM feature set

Accuracy: 86.06% RF With 300 trees Moments Feature Set

slide-35
SLIDE 35

Centre of Robotics

Execution Time

5 10 15 20 25 30 35 Wavelet Transform with blur detection Moments GLCM features Execution time in seconds Linear SVM kernel RBF SVM kernel Random forests

Wavelet Transform Less feature extraction Time

slide-36
SLIDE 36

Centre of Robotics

Conclusion

  • Strengths
  • Different densities of grasses
  • Multiple backgrounds (dry soil, muddy

soil, straw/stalk)

  • Grasses are a common weed in other

crops such as cotton.

  • Limitations
  • First technique dependents on growth

stage

  • Threshold of dilation, area removal

needs to determined.

  • Limited to a single type of weed
slide-37
SLIDE 37

Centre of Robotics

Topic

Plants Classification using Hough Line Transform & Support Vector Machine(SVM)

Researcher: Umar Muzaffar

52

slide-38
SLIDE 38

Centre of Robotics

Tools & Technology

  • Visual Studio
  • Image Processing( Opencv, C++)

53

slide-39
SLIDE 39

Centre of Robotics

Data set Collection

  • All dataset collected from:

 University of Central Punjab  Fields  Nurseries

54

slide-40
SLIDE 40

Centre of Robotics

Sample Images

Kangi Palm’s Plant Potato’s Plant Pea’s Plant (Captured from UCP) (Captured from Fields) (Captured from Nursery)

55

slide-41
SLIDE 41

Centre of Robotics

Training of data

  • There were total of 9 species which I

classified successfully

  • There were total of 300 images collected
  • Each specie consist of 33 images.
  • 31 images were used for testing purpose
  • 2 images were used for validation purpose

56

slide-42
SLIDE 42

Centre of Robotics

Techniques Used

  • Hough Line Transform (To extract

different shapes)

  • SVM (Support Vector Machine)

57

slide-43
SLIDE 43

Centre of Robotics

Flow Chart

58 Input Image Apply Hough Line Transform. Apply Canny Edge Detector Apply Bilateral Filter to reduce noise Find different shapes

  • f leaves

Apply SVM for classification If image’s data matches Output plant’s name Save Features in file If image’s data doesn’t match Output “It’s not match to existing data” Extract length & width

  • f leaves

Start End

slide-44
SLIDE 44

Centre of Robotics

Results

Cherry’s Plant

59

slide-45
SLIDE 45

Centre of Robotics

Continue. Cauliflower’s Plant

60

slide-46
SLIDE 46

Centre of Robotics

Continue. RedChilli’s Plant

61

slide-47
SLIDE 47

Centre of Robotics

Continue. Potato’s Plant

62

slide-48
SLIDE 48

Centre of Robotics

Continue. It’s Wall Palm Tree

63

slide-49
SLIDE 49

Centre of Robotics

Future Work

  • To improve my system, I will use different

techniques Like Odd Gabor Filters and morphological operations

  • It will help me to detect even veins of the

leaves

  • It will give much accurate results than, by

detecting the shapes of the leaves.

64

slide-50
SLIDE 50

Centre of Robotics

Disease Identification in Crops

Researcher: Sibtain Abbas

slide-51
SLIDE 51

Centre of Robotics

Goals

  • Increase in production.
  • Quality crops.
  • Reduce economic damage.
slide-52
SLIDE 52

Centre of Robotics

Losses in Punjab

Crop Value of Damage ($ millions) Cost of Control ($ millions) Rice 1.77 0.61 Wheat 1.83 0.40 Cotton 2.23 1.7 Totals 5.83 2.71

http://www.fin ance.gov.pk/survey/chapters_15/Annex_III_disease_damage.pdf

slide-53
SLIDE 53

Centre of Robotics

Common Diseases

  • Fusarium
  • Leaf Rust
  • Leaf Blotch
  • Wilt
  • Chlorosis
  • Scorch
slide-54
SLIDE 54

Centre of Robotics

Fusarium Blight

  • Fungal Disease
  • Causes
  • Effect on US Economy

http://www.ars.usda.gov/is/pr/2010/100401.htm

slide-55
SLIDE 55

Centre of Robotics

Leaf Scorch

  • Browning of Leaf Tissues, Veins and Tips.
  • Causes
  • Effect
slide-56
SLIDE 56

Centre of Robotics

Basic Steps

slide-57
SLIDE 57

Centre of Robotics

Flow Chart

slide-58
SLIDE 58

Centre of Robotics

Histogram- Methodology

Blurring the image.

Blurred Image

slide-59
SLIDE 59

Centre of Robotics

Histogram

  • HSV is used to improve color space

accuracy.

slide-60
SLIDE 60

Centre of Robotics

Histogram

  • Canny Edge Detection is used to further

enhance the details.

slide-61
SLIDE 61

Centre of Robotics

Histogram

  • Healthy and Diseased Histograms.
slide-62
SLIDE 62

Centre of Robotics

Histogram- Results

slide-63
SLIDE 63

Centre of Robotics

Multi Class SVM

  • Converting RGB to Gray Scale
  • Image Pre Processing
  • Image Segmentation
  • Feature Extraction
  • Classification
  • Testing
slide-64
SLIDE 64

Centre of Robotics

Multi Class SVM- Results

slide-65
SLIDE 65

Centre of Robotics

Multi Class SVM- Results

slide-66
SLIDE 66

Centre of Robotics

Multi Class SVM- Results

Stage No of Images Execution Time (sec) Feature Extraction 100 90 Training 25/ per class 3.3 Testing 20/ per class 0.7

slide-67
SLIDE 67

Centre of Robotics

Accuracy

  • Maximum accuracy achieved after 500

iterations.

slide-68
SLIDE 68

Centre of Robotics

Future Work

  • Improve the Accuracy.
  • Parallel detection of Weeds and Diseases.
slide-69
SLIDE 69

Centre of Robotics

Thank You!