The frequent occurrence of crop pests and diseases is one of the important factors leading to the reduction of crop quality and yield. Since pests are characterized by high similarity and fast movement, this poses a challenge for artificial intelligence techniques to identify pests in a timely and accurate manner. Therefore, we propose a new high-precision and real-time method for maize pest detection, Maize-YOLO. The network is based on YOLOv7 with the insertion of the CSPResNeXt-50 module and VoVGSCSP module. It can improve network detection accuracy and detection speed while reducing the computational effort of the model. We evaluated the performance of Maize-YOLO in a typical large-scale pest dataset IP102. We trained and tested against those pest species that are more damaging to maize, including 4533 images and 13 classes. The experimental results show that our method outperforms the current state-of-the-art YOLO family of object detection algorithms and achieves suitable performance at 76.3% mAP and 77.3% recall. The method can provide accurate and real-time pest detection and identification for maize crops, enabling highly accurate end-to-end pest detection.

Phenotyping provides important support for corn breeding. Unfortunately, the rapid detection of phenotypes has been the major limiting factor in estimating and predicting the outcomes of breeding programs. This study was focused on the potential of phenotyping to support corn breeding using unmanned aerial vehicle (UAV) images, aiming at mining and deepening UAV techniques for comparing phenotypes and screening new corn varieties. Two geometric traits (plant height, canopy leaf area index (LAI)) and one lodging resistance trait (lodging area) were estimated in this study. It was found that stereoscopic and photogrammetric methods were promising ways to calculate a digital surface model (DSM) for estimating corn plant height from UAV images, with R² = 0.7833 (p < 0.001) and a root mean square error (RMSE) = 0.1677. In addition to a height estimation, the height variation was analyzed for depicting and validating the corn canopy uniformity stability for different varieties. For the lodging area estimation, the normalized DSM (nDSM) method was more promising than the gray-level co-occurrence matrix (GLCM) textural features method. The estimation error using the nDSM ranged from 0.8% to 5.3%, and the estimation error using the GLCM ranged from 10.0% to 16.2%. Associations between the height estimation and lodging area estimation were done to find the corn varieties with optimal plant heights and lodging resistance. For the LAI estimation, the physical radiative transfer PROSAIL model offered both an accurate and robust estimation performance both at the middle (R² = 0.7490, RMSE = 0.3443) and later growing stages (R² = 0.7450, RMSE = 0.3154). What was more exciting was that the estimated sequential time series LAIs revealed a corn variety with poor resistance to lodging in a study area of Baogaofeng Farm. Overall, UAVs appear to provide a promising method to support phenotyping for crop breeding, and the phenotyping of corn breeding in this study validated this application.

Seed maize and common maize plots have different planting patterns and variety types. Identification of seed maize is the basis for seed maize growth monitoring, seed quality and common maize seed supply. In this paper, a random forest (RF) classifier is used to develop an approach for seed maize fields’ identification, using the time series vegetation indexes (VIs) calculated from multispectral data acquired from Landsat 8 and Gaofen 1 satellite (GF-1), field sample data, and texture features of Gaofen 2 satellite (GF-2) panchromatic data. Huocheng and Hutubi County in the Xinjiang Uygur Autonomous Region of China were chosen as study area. The results show that RF performs well with the combination of six VIs (normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), triangle vegetation index (TVI), ratio vegetation index (RVI), normalized difference water index (NDWI) and difference vegetation index (DVI)) and texture features based on a grey-level co-occurrence matrix. The classification based on “spectrum + texture” information has higher overall, user and producer accuracies than that of spectral information alone. Using the “spectrum + texture” method, the overall accuracy of classification in Huocheng County is 95.90%, the Kappa coefficient is 0.92, and the producer accuracy for seed maize fields is 93.91%. The overall accuracy of the classification in Hutubi County is 97.79%, the Kappa coefficient is 0.95, and the producer accuracy for seed maize fields is 97.65%. Therefore, RF classifier inputted with high-resolution remote-sensing image features can distinguish two kinds of planting patterns (seed and common) and varieties types (inbred and hybrid) of maize and can be used to identify and map a wide range of seed maize fields. However, this method requires a large amount of sample data, so how to effectively use and improve it in areas lacking samples needs further research.

The tassel development status and its branch number in maize flowering stage are the key phenotypic traits to determine the growth process, pollen quantity of different maize varieties, and detasseling arrangement for seed maize production fields. Rapid and accurate detection of tassels is of great significance for maize breeding and seed production. However, due to the complex planting environment in the field, such as unsynchronized growth stage and tassels vary in size and shape caused by varieties, the detection of maize tassel remains challenging problem, and the existing methods also cannot distinguish the early tassels. In this study, based on the time series unmanned aerial vehicle (UAV) RGB images with maize flowering stage, we proposed an algorithm for automatic detection of maize tassels which is suitable for complex scenes by using random forest (RF) and VGG16. First, the RF was used to segment UAV images into tassel regions and non-tassel regions, and then extracted the potential tassel region proposals by morphological method; afterwards, false positives were removed through VGG16 network with the ratio of training set to validation set was 7:3. To demonstrate the performance of the proposed method, 50 plots were selected from UAV images randomly. The precision, recall rate and F1-score were 0.904, 0.979 and 0.94 respectively; 50 plots were divided into early, middle and late tasseling stages according to the proportion of tasseling plants and the morphology of tassels. The result of tassels detection was late tasseling stage > middle tasseling stage > early tasseling stage, and the corresponding F1-score were 0.962, 0.914 and 0.863, respectively. It was found that the model error mainly comes from the recognition of leaves vein and reflective leaves as tassels. Finally, to show the morphological characteristics of tassel directly, we proposed an endpoint detection method based on the tassel skeleton, and further extracted the tassel branch number. The method proposed in this paper can well detect tassels of different development stages, and support large scale tassels detection and branch number extraction.

The accurate and timely access to the spatial distribution information of crops is of great importance for agricultural production management. Although widely used, supervised classification mapping requires a large number of field samples, and is consequently costly in terms of time and money. In order to reduce the need for sample size, this paper proposes an unsupervised classification method based on principal components isometric binning (PCIB). In particular, principal component analysis (PCA) dimensionality reduction is applied to the classification features, followed by the division of the top k principal components into equidistant bins. Bins of the same category are subsequently merged as a class label. Multitemporal Gaofen 1 satellite (GF-1) remote sensing images were collected over the southwest of Hulin City and Luobei County of Hegang City, Heilongjiang Province, China in order to map crop types in 2016 and 2017. Our proposed method was compared with commonly used classifiers (random forest, K-means and Iterative Self-Organizing Data Analysis Techniques Algorithm (ISODATA)). Results demonstrate PCIB and random forest to have the highest classification accuracies, reaching 82% in 2016 in the southwest of Hulin City. In Luobei County in 2016, the accuracies of PCIB and random forest were determined as 81% and 82%, respectively. It can be concluded that the overall accuracy of our proposed method meets the basic requirements of classification accuracy. Despite exhibiting a lower accuracy than that of random forest, PCIB does not require a large field sample size, thus making it more suitable for large-scale crop mapping.

Accurate, year-by-year crop distribution information is a key element in agricultural production regulation and global change governance. However, due to the high sampling costs and insufficient use of historical samples, a supervised classifying method for sampling every year is unsustainable for mapping crop types over time. Therefore, this paper proposes a method for the generation and screening of new samples for 2018 based on historical crop samples, and then it builds a crop mapping model for that current season. Pixels with the same crop type in the historical year (2013–2017) were extracted as potential samples, and their spectral features and spatial information in the current year (2018) were used to generate new samples based on clustering screening. The research result shows that when the clustering number is different, the number and structure of new generated sample also changes. The sample structure generated in Luobei County was not balanced, with the ‘other crop’ representing less than 3.97%, but the structure of southwest Hulin City was more balanced. Based on the newly generated samples and the ground reference data of classified year, the classification models were constructed. The average classification accuracies of Luobei County in 2018 based on new generated samples and field samples were 69.35% and 77.59%, respectively, while those of southwest Hulin City were 80.44% and 82.94%, respectively. Combined with historical samples and the spectral information of the current year, this study proposes a method to generate new samples. It can overcome the problem of crop samples only being collected in the field due to the difficulty of visual interpretation, effectively improve the use of historical data, and also provide a new idea for sustainable crop mapping in many regions lacking seasonal field samples.