Imaging in agriculture is where machine learning will be necessary to execute on a chip. Using triangulation as an example for generating point cloud data at high resolution would require a tremendous amount of computation. Supercomputers based on GPU might help but that won’t make it into the field devices, therefore, I am predicting that a new class of ASIC will emerge that do very specialized image processing or field level devices.
Autonomous navigation based on stereo vision was successfully achieved in several research studies using different cues. For example, in crop rows, cut-uncut edges, ridges, furrows, artificial markers, swaths and even stubble can be used. Kise et al.
[38] developed a stereo vision system that uses the 3-D crop row structure for automated guidance; problems like high computational load and blank pixels of some locations (particularly the ones that are further away) were reported, but were addressed by usingSource: 3-D Imaging Systems for Agricultural Applications—A Review
