arXiv subjects

Yunfeng Nie

Publications and source records attributed to Yunfeng Nie.

4 recordsLinked to original sources

arXiv2026

Task-Driven Lens Design

Classical lens design minimizes optical aberrations to produce sharp images, but is typically decoupled from downstream computer vision tasks. Existing end-to-end optical design learns optical encoding through joint optimization, but often suffers from an unstable training process. We propose task-driven lens design, a new optimization philosophy for joint optics-network systems. We freeze the pretrained vision model and optimize only the lens so that the image formation better fits the model's feature preferences. This network-frozen setting yields a low-dimensional and stable optimization process, enabling lens design from scratch without human intervention, thereby exploring a broader design space. Multiple computer vision experiments show that TaskLenses outperform classical ImagingLenses with the same or even fewer elements. Our analysis reveals that the learned optics exhibit long-tailed point spread functions, better preserving preferred structural cues when aberrations cannot be fully corrected. These results highlight task-driven design as a practical route for optical lenses that are compatible with modern vision models, and also inspire new optical design objectives beyond traditional aberration minimization.

cs.CV↗

arXiv2024

Surface plasmon resonance sensor based on a novel prism for the detection of a broad range of polymers

Microplastics are pervasive pollutants that pose threats to the environment and human health. Detecting them reliably represents a crucial step toward finding solutions. While traditional techniques like vibrational and mass spectroscopies are time-consuming and have limitations in analysing polymers with pigment additives, surface plasmon resonance (SPR) shows promise due to its high sensitivity and rapid detection capabilities.However, existing SPR methods rely on one output variable and fail to determine various microplastic parameters simultaneously. To address this uncertainty between size and polymer type, we propose a prism design that enables SPR resonance in a refractive index range of 1.402-1.56 RIU (Refractive Index Units) for polymer detection. Experimental verification of the prism's dynamic range using refractive index oils presented an agreement of 98.47% with theoretical analysis. Furthermore, we acquired SPR curves from ten different polymer types, including Nitrile, Viton and Tyre rubber. These results demonstrate SPR's potential to identify polymers and its ability to detect even pigmented materials, often challenging for other techniques. The imaging capabilities of the novel prism are presented, indicating its potential for future imaging of microplastics. Our approach holds promise for advancing microplastics analysis by enabling comprehensive assessment of particle parameters, such as polymer type, size, quantity and shape, and expanding the range of detectable polymer types.

physics.ins-det↗

arXiv2022

A survey on computational spectral reconstruction methods from RGB to hyperspectral imaging

Hyperspectral imaging enables versatile applications due to its competence in capturing abundant spatial and spectral information, which are crucial for identifying substances. However, the devices for acquiring hyperspectral images are expensive and complicated. Therefore, many alternative spectral imaging methods have been proposed by directly reconstructing the hyperspectral information from lower-cost, more available RGB images. We present a thorough investigation of these state-of-the-art spectral reconstruction methods from the widespread RGB images. A systematic study and comparison of more than 25 methods has revealed that most of the data-driven deep learning methods are superior to prior-based methods in terms of reconstruction accuracy and quality despite lower speeds. This comprehensive review can serve as a fruitful reference source for peer researchers, thus further inspiring future development directions in related domains.

eess.IV↗

arXiv2020

Freeform imaging system design with multiple reflection surfaces

Reflective imaging systems form an important part of photonic devices such as spectrometers, telescopes, augmented and virtual reality headsets or lithography platforms. Reflective optics provide unparalleled spectral performance and can be used to reduce overall volume and weight. So far, most reflective designs have focused on two or three reflections, while four-reflection freeform designs can deliver a higher light throughput (faster F-number) as well as a larger field-of-view (FOV). However, advanced optical design strategies for four-reflection freeform systems have been rarely reported in literature. This is due to the increased complexity in solution space but also the fact that additional mirrors hinder a cost-effective realization (manufacture, alignment, etc.). Recently, we have proposed a novel design method to directly calculate the freeform surface coefficients while merely knowing the mirror positions and tilts. Consequently, this method allows laymen with basic optical design knowledge to calculate 'first time right' freeform imaging systems in a matter of minutes. This contrasts with most common freeform design processes, which requires considerable experience, intuition or guesswork. Firstly, we demonstrate the effectiveness of the proposed method for a four-mirror high-throughput telescope with 250mm-focal-length, F/2.5 and a wide rectangular FOV of 8.5° x 25.5°. In a subsequent step, we propose an effective three-mirror but four-reflection imaging system, which consists of two freeform mirrors and one double-reflection spherical mirror. Compared with common three-mirror and three-reflection imagers, our novel multi-reflection system shows unprecedented possibilities for an economic implementation while drastically reducing the overall volume.

physics.optics↗