# Machine vision-based detection of browning maturity in shiitake cultivation sticks

## TL;DR

This paper introduces a machine vision system to detect browning maturity in shiitake cultivation sticks, improving monitoring and quality control.

## Contribution

A novel two-stage image segmentation approach combining VG-Stick-YOLOv11 and RS-UNet for accurate and efficient browning maturity detection.

## Key findings

- VG-Stick-YOLOv11 achieved 95.80% mIoU for contour extraction with reduced parameters and computation.
- RS-UNet achieved 94.35% segmentation accuracy and 88.56% IoU for browning regions.
- RS-UNet reduced parameters by 36.31% compared to the ResNet50-U-Net baseline.

## Abstract

Accurate monitoring of pigmentation changes during the browning stage of shiitake cultivation sticks is essential for assessing substrate maturity, forecasting mushroom emergence, and improving cultivation quality. However, current commercial detection methods lack objective, real-time, and quantifiable evaluation indicators for assessing the browning degree.

This study proposes a two-stage image segmentation approach to address this issue. First, a novel VG-Stick-YOLOv11 model, built upon YOLOv11n-seg with VanillaNetBlock and GhostConv, was developed for real-time contour extraction and browning stage classification of shiitake sticks. Based on the extracted features, machine learning techniques facilitated rapid, semi-automatic annotation of browning regions, thereby constructing a segmentation dataset. Finally, the ResNet-Stick-UNet (RS-UNet) model was designed for precise browning region segmentation and area ratio calculation. The encoder utilizes ResNet50 with multi-branch inputs and stacked small kernels to enhance feature extraction, while the decoder incorporates a hybrid structure of grouped and depthwise separable convolutions for efficient channel fusion and detail preservation. A spatial attention mechanism was embedded in skip connections to emphasize large-scale browning regions.

The proposed VG-Stick-YOLOv11 achieved the best mIoU of 95.80% for stick contour extraction while markedly reducing parameters and computation. For browning region segmentation, RS-UNet achieved a high segmentation accuracy of 94.35% and an IoU of 88.56%, outperforming comparison models such as Deeplabv3+ and Swin-UNet. Furthermore, RS-UNet reduced the number of parameters by 36.31% compared to the ResNet50-U-Net baseline.

The collaborative two-stage model provides an effective and quantitative solution for maturity detection of shiitake cultivation sticks during the browning stage. This work promotes the intelligent and standardized development of shiitake substrate cultivation.