Table 4.

Summary of the technical difficulties of the 3-D techniques used in agricultural applications.

Basic Principle	Technique	Application	Technical Difficulties
Triangulation	Stereo vision	-Autonomous navigation [38,39,40,42,44,46] -Crop husbandry [71,98,100] -Animal husbandry [121,132]	-Blank pixels of some locations specially the ones that are further away from the camera -Low light (cloudy sky) affects 3-D point generation -Direct sunlight and shadows in a sunny day affect strongly the depth image generation -Uniform texture of long leaves affect the 3-D point generation -Limited field of view -External illumination is required for night implementations -Correspondence and parallax problems -A robust disparity estimation is difficult in areas of homogeneous colour or occlusion -Specular reflections -Colour heterogeneity of the target object -A constant altitude needs to be maintained if a stereo vision system is mounted on a UAV -Camera calibration is necessary -Occlusion of leaves -Selection of a suitable camera position
	Multi-view stereo	-Crop husbandry [65] -Animal husbandry [124]	-Surface integration from multiple views is the main obstacle -Challenging software engineering if high-resolution surface reconstruction is desired -Software obstacles associated with handling large images during system calibration and stereo matching
	Multiple-baseline stereo	-Autonomous navigation [43]	-Handling a rich 3-D data is computationally demanding
	Structure-from-motion	-Crop husbandry [64,67,68,69,97]	-Occlusion of leaves -Plant changing position from one image to the other due to the wind -High computation power is required to generate a dense point cloud -Determination of a suitable Image overlapping percentage -Greater hectare coverage requires higher altitudes when using UAVs -The camera’s pixel resolution determines the field spatial resolution -Image mosaicking is technically difficult from UAVs due to the translational and rotational movements of the camera
	Shape-from-Silhouette	-Crop husbandry [87,88,89]	-3-D reconstruction results strongly depend on good image pre-processing -Camera calibration is important if several cameras are used -Dense and random canopy branching is more difficult to reconstruct -Post-processing filtering may be required to remove noisy regions
	Structured light (light volume) sequentially coded	-Crop husbandry [93]	-Limited projector depth of field -High dynamic range scene -Internal reflections -Thin objects -Occlusions
	Structured light (light volume) pseudo random pattern	-Autonomous navigation [47] -Animal husbandry [122,128]	-Strong sensitivity to natural light -Small field of view -Smooth and shiny surfaces do not produce reliable depth measurements -Misalignment between the RGB and depth image due to the difference in pixel resolution -Time delay (30 s) for a stable depth measurement after a quick rotation -Mismatch between the RGB and depth images’ field of view and point of view
	Shape-from-Shading	-Crop husbandry [101]	-A zigzag effect at the target object’s boundary is generated (in interlaced video) if it moves at high speeds
	Structured light shadow Moiré	-Crop husbandry [94]	-Sensitive to disturbances (e.g., surface reflectivity) that become a source of noise
	Shape-from-focus	-Crop husbandry [90]	-Limited depth of field decreases the accuracy of the 3-D reconstruction
TOF	Pulse modulation (light sheet)	-Autonomous navigation [49] -Crop husbandry [106,109]	-Limited perception of the surrounding structures -Requires movement to obtain 3-D data -Pitching, rolling or jawing using servo motors (i.e., pan-tilt unit) is a method to extend the field of view, but adds technical difficulties -Point cloud registration requires sensor fusion -Small plants are difficult to detect -Lower sampling rate and accuracy compared to continuous wave modulation TOF
	Pulse modulation (light volume)	-Autonomous navigation and crop husbandry [107]	-Limited pixel resolution -Difficulty to distinguish small structures with complex shapes
	Continuous wave modulation (light sheet)	-Crop husbandry [109]	-Poor distance range measurement (up to 3 m)
	Continuous wave modulation (light volume)	-Crop husbandry [110,111,112] -Animal husbandry [122]	-Small field of view -Low pixel resolution -Calibration could be required to correct radial distortion -Requires a sunlight cover for better results -Limited visibility due to occlusion -Lack of colour output that could be useful for a better image segmentation
Inter-ferometry	White-light	-Crop husbandry [115,116,120]	-The scattering surface of the plant forms speckles that affect the accuracy -Complexity of implementation
	Holographic	-Crop husbandry [119]	-Need of a reference object in the image to detect disturbances
	Speckle	-Crop husbandry [117,118]	-Agricultural products with rough surface could be difficult to reconstruct -High camera resolutions provide better capabilities to resolve high fringe densities