Weed control is fundamental to modern agriculture, underpinning crop productivity, food security, and the economic sustainability of farming operations. Herbicides have long been the cornerstone of effective weed management, significantly enhancing agricultural yields over recent decades. However, the field now faces critical challenges, including stagnation in the discovery of new herbicide modes of action (MOAs) and the escalating prevalence of herbicide-resistant weed populations. High research and development costs, coupled with stringent regulatory hurdles, have impeded the introduction of novel herbicides, while the widespread reliance on glyphosate-based systems has accelerated resistance development. In response to these issues, advanced image-based plant phenotyping technologies have emerged as pivotal tools in addressing herbicide-related challenges in weed science. Utilizing sensor technologies such as hyperspectral, multispectral, RGB, fluorescence, and thermal imaging methods, plant phenotyping enables the precise monitoring of herbicide drift, analysis of resistance mechanisms, and development of new herbicides with innovative MOAs. The integration of machine learning algorithms with imaging data further enhances the ability to detect subtle phenotypic changes, predict herbicide resistance, and facilitate timely interventions. This review comprehensively examines the application of image phenotyping technologies in weed science, detailing various sensor types and deployment platforms, exploring modeling methods, and highlighting unique findings and innovative applications. Additionally, it addresses current limitations and proposes future research directions, emphasizing the significant contributions of phenotyping advancements to sustainable and effective weed management strategies. By leveraging these sophisticated technologies, the agricultural sector can overcome existing herbicide challenges, ensuring continued productivity and resilience in the face of evolving weed pressures.

Image-based spectroscopy phenotyping is a rapidly growing field that investigates how genotype, environment and management interact using remote or proximal sensing systems to capture images of a plant under multiple wavelengths of light. While remote sensing techniques have proven effective in crop phenotyping, they can be subject to various noise sources, such as varying lighting conditions and plant physiological status, including leaf orientation. Moreover, current proximal leaf-scale imaging devices require the sensors to accommodate the state of the samples during imaging which induced extra time and labor cost. Therefore, this study developed a proximal multispectral imaging device that can actively attract the leaf to the sensing area (target-to-sensor mode) for high-precision and high-throughput leaf-scale phenotyping. To increase the throughput and to optimize imaging results, this device innovatively uses active airflow to reposition and flatten the soybean leaf. This novel mechanism redefines the traditional sensor-to-target mode and has relieved the device operator from the labor of capturing and holding the leaf, resulting in a five-fold increase in imaging speed compared to conventional proximal whole leaf imaging device. Besides, this device uses artificial lights to create stable and consistent lighting conditions to further improve the quality of the images. Furthermore, the touch-based imaging device takes full advantage of proximal sensing by providing ultra-high spatial resolution and quality of each pixel by blocking the noises induced by ambient lighting variances. The images captured by this device have been tested in the field and proven effective. Specifically, it has successfully identified nitrogen deficiency treatment at an earlier stage than a typical remote sensing system. The p-value of the data collected by the device (p = 0.008) is significantly lower than that of a remote sensing system (p = 0.239).

Western corn rootworm (WCR) is one of the most devastating corn rootworm species in North America because of its ability to cause severe production loss and grain quality damage. To control the loss, it is important to identify the infection of WCR at an early stage. Because the root system is the earliest feeding source of the WCR at the larvae stage, assessing the direct damage in the root system is crucial to achieving early detection. Most of the current methods still necessitate uprooting the entire plant, which could cause permanent destruction and a loss of the original root’s structural information. To measure the root damages caused by WCR non-destructively, this study utilized MISIRoot, a minimally invasive and in situ automatic plant root phenotyping robot to collect not only high-resolution images but also 3D positions of the roots without uprooting. To identify roots in the images and to study how the damages were distributed in different types of roots, a deep convolution neural network model was trained to differentiate the relatively thick and thin roots. In addition, a color camera was used to capture the above-ground morphological features, such as the leaf color, plant height, and side-view leaf area. To check if the plant shoot had any visible symptoms in the inoculated group compared to the control group, several vegetation indices were calculated based on the RGB color. Additionally, the shoot morphological features were fed into a PLS-DA model to differentiate the two groups. Results showed that none of the above-ground features or models output a statistically significant difference between the two groups at the 95% confidence level. On the contrary, many of the root structural features measured using MISIRoot could successfully differentiate the two groups with the smallest t-test p-value of 1.5791 × 10⁻⁶. The promising outcomes were solid proof of the effectiveness of MISIRoot as a potential solution for identifying WCR infestations before the plant shoot showed significant symptoms.

Spectral imaging has been widely applied in plant phenotyping to assess corn leaf nitrogen status. Recent studies indicate that spatial variations within a single leaf’s multispectral image provide stronger signals for corn nitrogen estimation. However, current technologies for corn multispectral imaging cannot capture a large corn leaf segment with high-resolution and simple operation, limiting their efficiency and accuracy in nitrogen estimation. To address this gap, this study developed a proximal multispectral imaging device that can capture high-resolution snapshot multispectral images of a large segment of a single corn leaf. This device uses airflow to autonomously position and flatten the leaf to minimize the noise in images due to leaf curvature and simplify operation. Moreover, this device adopts a transmittance imaging regime by clamping the corn leaf between the camera and the lighting source to block the environmental lights and supply uniform lighting to capture high-resolution and high-precision leaf images within six seconds. A field assay was conducted to validate the effectiveness of the multispectral images captured by this device in assessing nitrogen status by classifying the nitrogen treatments applied to corn. Six nitrogen treatments were applied to 12 plots of corn fields, and 10 images were collected at each plot. By using the average vegetative index of the whole image, only one treatment was significantly different from the other five treatments, and no significant difference was observed among any other groups. However, by extracting the spatial and spectral features from the images and combining these features, the accuracy of nitrogen treatment classification improved compared to using the average index. In another analysis, by applying spatial–spectral analysis methods to the images, the nitrogen treatment classification accuracy has improved compared to using the average index. These results demonstrated the advantages of this high-resolution and high-throughput imaging device for distinguishing nitrogen treatments by facilitating spatial–spectral combined analysis for more precise classification.

Weed control is fundamental to modern agriculture, underpinning crop productivity, food security, and the economic sustainability of farming operations. Herbicides have long been the cornerstone of effective weed management, significantly enhancing agricultural yields over recent decades. However, the field now faces critical challenges, including stagnation in the discovery of new herbicide modes of action (MOAs) and the escalating prevalence of herbicide-resistant weed populations. High research and development costs, coupled with stringent regulatory hurdles, have impeded the introduction of novel herbicides, while the widespread reliance on glyphosate-based systems has accelerated resistance development. In response to these issues, advanced image-based plant phenotyping technologies have emerged as pivotal tools in addressing herbicide-related challenges in weed science. Utilizing sensor technologies such as hyperspectral, multispectral, RGB, fluorescence, and thermal imaging, plant phenotyping enables precise monitoring of herbicide drift, analysis of resistance mechanisms, and development of new herbicides with innovative MOAs. The integration of machine learning algorithms with imaging data further enhances the ability to detect subtle phenotypic changes, predict herbicide resistance, and facilitate timely interventions. This review comprehensively examines the application of image phenotyping technologies in weed science, detailing various sensor types and deployment platforms, exploring modeling methods, and highlighting unique findings and innovative applications. Additionally, it addresses current limitations and proposes future research directions, emphasizing the significant contributions of phenotyping advancements to sustainable and effective weed management strategies. By leveraging these sophisticated technologies, the agricultural sector can overcome existing herbicide challenges, ensuring continued productivity and resilience in the face of evolving weed pressures.