As materials science has extended into the realm of nanometers, the ability to observe and analyze nanoscale features has become increasingly crucial. However, because of the existence of the diffraction limit, characterizing optical properties of nanostructures is challenging. Traditional near-field scanning optical microscopy (NSOM) has been developed to break the diffraction limit, but it still has some disadvantages. For example, aperture NSOM has low signal levels and limited spatial resolution, and apertureless (scattering-type) NSOM requires extra interferometry to exterminate background signals, obstructing broadband spectrum measurements. Recently, background-free NSOM based on surface plasmon polariton (SPP) nanofocusing has been developed and higher spatial resolution and signal-to-noise ratio has been demonstrated. However, present nanofocusing NSOM still has some challenges such as efficiently coupling light into SPP and obtaining more cut-off-free SPP modes that contribute to the nanoscale resolution. This dissertation presents a novel photonic-plasmonic probe for near-field optical nanoimaging, which selectively couples the light in a single-mode optical fiber into the SPP on a silver nanowire waveguide, and adiabatically nanofocuses the SPP to create a nanoscale light source. The design of this coupler probe is assisted by deep learning and the coupling efficiency is validated by experiments. This dissertation covers four parts:1. Deep Learning Assisted Inverse Design of the Photonic-Plasmonic Coupler 2. Deterministic assembling the Fluorescent Nanodiamonds to the Photonic-Plasmonic Probe for Enhanced Signals in Magnetic Field Sensing 3. Hyperspectral Nanoimaging Accidental Quasi-Bound States in the Continuum in a Single Silver Nanocube by the Photonic-Plasmonic Probe 4. Imaging Near-field Optical Chirality and Deep Learning Accelerated Chirality Prediction.