Abstract:Weed identification in vegetable plantation is more challenging than crop weed identification due to their random plant spacing. So far, little work has been found on identifying weeds in vegetable plantation. Traditional methods of crop weed identification used to be mainly focused on identifying weed directly; however, there is a large variation in weed species. This paper proposes a new method in a contrary way, which combines deep learning and image processing technology. Firstly, a trained CenterNet model was used to detect vegetables and draw bounding boxes around them. Afterwards, the remaining green objects falling out of bounding boxes were considered as weeds. In this way, the model focuses on identifying only the vegetables and thus avoid handling various weed species. Furthermore, this strategy can largely reduce the size of training image dataset as well as the complexity of weed detection, thereby enhancing the weed identification performance and accuracy. To extract weeds from the background, a color index-based segmentation was performed utilizing image processing. The employed color index was determined and evaluated through Genetic Algorithms (GAs) according to Bayesian classification error. During the field test, the trained CenterNet model achieved a precision of 95.6%, a recall of 95.0%, and a $F_{1}$ score of 0.953, respectively. The proposed index −19R + 24G −2B ≥ 862 yields high segmentation quality with a much lower computational cost compared to the wildly used ExG index. These experiment results demonstrate the feasibility of using the proposed method for the ground-based weed identification in vegetable plantation.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve This paper aims to address the issue of weed identification in vegetable plantations. Specifically, the research proposes a new method that combines deep learning and image processing techniques to identify weeds in vegetable plantations. #### Main Objectives - Develop a weed identification algorithm based on deep learning and image processing for robots to remove weeds in vegetable plantations. #### Specific Objectives 1. Train a model using deep learning methods to detect vegetables (using cabbage as an example) and draw bounding boxes. 2. Use image processing techniques to extract and segment green objects outside the bounding boxes, i.e., weeds. ### Research Background - Weed identification in vegetable plantations is more challenging than crop weed identification because the planting spacing of vegetables and weeds is randomly distributed. - Traditional weed identification methods mainly focus on directly identifying weeds, but due to the vast variety of weeds, this approach has limitations. - This study proposes an opposite approach, which first identifies and segments vegetables, then considers the remaining green objects as weeds. This can reduce the size of the training image dataset and the complexity of weed detection, thereby improving the performance and accuracy of weed identification. ### Method Overview 1. **Deep Learning Model**: Use the CenterNet model for vegetable detection and draw bounding boxes. 2. **Image Processing**: Use color indices to segment green objects outside the bounding boxes and identify these objects as weeds. 3. **Genetic Algorithm**: Optimize the color indices using a genetic algorithm to reduce computational costs and improve segmentation quality. ### Experimental Results - The CenterNet model achieved an accuracy of 95.6% on the test set, with a recall rate of 95% and an F1 score of 0.953. - The proposed color index (-19R + 24G - 2B ≥ 862) has higher segmentation quality and lower computational costs compared to the commonly used ExG index. ### Conclusion Experimental results show that the proposed method effectively identifies weeds in vegetable plantations under natural conditions and is suitable for intelligent robotic weed control.

Weed Identification Using Deep Learning and Image Processing in Vegetable Plantation

A novel deep learning‐based method for detection of weeds in vegetables

Identification of maize and wheat seedlings and weeds based on deep learning

Weed Identification by Single-Stage and Two-Stage Neural Networks: A Study on the Impact of Image Resizers and Weights Optimization Algorithms

Weed detection by using image processing

Deep Learning-Based Crop Weed Recognition

Weed recognition using deep learning techniques on class-imbalanced imagery

Efficient Crop Segmentation Net and Novel Weed Detection Method

Evaluation of Two Deep Learning-Based Approaches for Detecting Weeds Growing in Cabbage

Weed25: A deep learning dataset for weed identification

High-information Data-Centric Weed Images Data Identification System Based on Triple Filter for Smart Agriculture

Semantic Segmentation for Weed Detection in Corn.

A deep learning-based method for classification, detection, and localization of weeds in turfgrass.

Towards weeds identification assistance through transfer learning

Weed Recognition at Soybean Seedling Stage Based on YOLOV8nGP + NExG Algorithm

Graph weeds net: A graph-based deep learning method for weed recognition

An Instance Segmentation-Based Method to Obtain the Leaf Age and Plant Centre of Weeds in Complex Field Environments

Deep Learning Techniques for In-Crop Weed Identification: A Review

Real-Time Crop Recognition In Transplanted Fields With Prominent Weed Growth: A Visual-Attention-Based Approach

Deep Learning-Based Object Detection System for Identifying Weeds Using UAS Imagery