Table 4 Part 1: an analytical study on examples of crop diseases protection and weeds detection approaches; highlights the applied algorithm, plant and data type, data pre-processing and the extracted features

[21]

Leaf disease recognition

CNN

Tea plant

Digital images

Data augmentation

/

[63]

Plant disease detection

CNN

Plant leaves

Digital images

Data augmentation

/

[9]

Plant disease detection

DNN

Plant leaves

Digital images

Data augmentation

[143]

Light leaf spotdetection

SVM

Soilseedrape

Multi-spectralimages

Removal of:background and redundant features

Carter Index 1 light leaf spot index Spectral signature

[82]

Disease detection

SVM

Wheat

Sensing data

/

NDVI, Photochemical reflection index Pigment-specific simple ratio, water index

[155]

Diseasedetection

CNN

Wheat

Remote sensingdata

Segmentaion: sliding-window

3D blocks

[125]

Leaf disease recognition

CNN

Maize

Digital images

/

RGB

[15]

Plant disease detection

Gated recurrent unit CNN

Soybean

Satellite images Crop rotation

Time-series

Spectral bands of: red, green, blue NIR, NDVI

[150]

Leaf disease detection

CNN

Tomato

Digital images

Data augmentation:resolution reducingbicubic method to enlarge images

Patches

[136]

Plant disease detection

RF

Wheat

Aerial multi-spectralimages

/

RVI, NDVI, OSAVI NIR, Red

[49]

Plant disease detection

Partial leastsquares regression

Wheat

Aerial hyper-spectralimages

Image fusion and mosaicking

Disease index many vegetation indexes texture features

[153]

Plant detection

CNN

Maize

Aerial RGBimages

Segmentation by RF

Morphological caracteristics of maize tassels

[80]

Plant disease detection

ANN

Wheat

Hyper-spectral aerial images

Fusion and stitchingradiometric calibrationatmospheric correction

11 Vegetation indexes spectral bands texture features

[88]

Plant disease detection

deep learning CNN: AlexNet GoogLeNet

Plant leaves

Digital images

Coloured, gray-scaled, segmented

/

[2]

Crop and weed classification

SVM RBF kernel

Chilli, Pigweed Marsh herb Lamb’s quarters Cogongrass, cucumber

Digital images

Segmentation: binarisation technique: -global threshold noise removal morphological opening and morphological closing

14 features: RGB colours, shape features Moment invariant features

[44]

Weeds detection

RF

Maize

Hyper-spectral images

Segmentation

185 features Ratio Vegetation Index,NDVI

[3]

Weeds detection

SVM Gaussian kernel

Corn leaves and broad silver beet leaves

Spectral reflectance images

/

NDVI

[8]

Crop and weed classification

ANN Generalised Softmax Perceptron and the Posterior Probability Model Selection algorithm

Sunflower

Digital images

Special process of segmentation

13 morphological features: Number of boundary pixelsCompactness, Perimeter, Centroid and Elongation,The geometric centre Area, Number of pixels of objects, Major and minor axis of the best fit ellipse