Fine-grained image classification methods often suffer from the challenge that the subordinate categories within an entry-level category can only be distinguished by subtle differences. Crop disease classification is affected by various visual interferences, including uneven illumination, dew, and equipment jitter. It demands an effective algorithm to accurately discriminate one category from the others. Thus, the representational ability of algorithm needs to be strengthened to learn a robust domain-specific discrimination through an effective way. To address this challenge, a unified convolutional neural network (CNN) denoting the matrix-based convolutional neural network (M-bCNN) was proposed. Its hallmark is the convolutional kernel matrix, whose convolutional layers are arranged parallelly in the form of a matrix, and integrated with DropConnect, exponential linear unit, local response normalization, and so on to defeat over-fitting and vanishing gradient. With a tolerable addition of parameters, it can effectively increase the data streams, neurons, and link channels of the model compared with the commonly used plain networks. Therefore, it will create more non-linear mappings and will enhance the representational ability with a tolerable growth of parameters. The images of winter wheat leaf diseases were utilized as experimental samples for their strong similarities among sub-categories. A total of 16652 images containing eight categories were collected from Shandong Province, China, and were augmented into 83260 images. The M-bCNN delivered significant improvements and achieved an average validation accuracy of 96.5% and a testing accuracy of 90.1%; this outperformed AlexNet and VGG-16. The M-bCNN demonstrated accuracy gains with a convolutional kernel matrix in fine-grained image classification.

Maturity level based classification system plays an essential role in the design of tomato harvesting robot. Traditional knowledge based systems are unable to meet the current production management requirements of precision picking, because they are time-consuming and have low accuracy. Our research proposes an improved deep learning based classification method that improves accuracy and scalability of tomato ripeness with a small amount of training data. This study was on the relationship of different dataset augmentation methods and prediction results of final classification task. We implemented classification systems based on convolutional neural network (CNN), by training and validating the model on different augmented datasets and tried to choose an optimal augmentation method for datasets. The experimental results showed an average accuracy of 91.9% with a less than 0.01 second prediction time. Compared to existing methods, our solution achieved better prediction results both in terms of accuracy and time consumption. Moreover, this is a versatile method and can be extended to other related fields.

China's soybean supply and demand are seriously imbalanced. It is crucial to improve the level of soybean breeding. Hundred-grain weight is one of the most essential phenotypic parameters for crop breeding. Accurate soybean seed counting is a key step for 100-grain weight. There are several seed counting methods, which have their own limitations one way or the other. Among these, manual counting is time-consuming, electronic automatic seed counter devices are expensive and their counting speed is very slow, and the traditional digital image processing techniques are not suitable for seed counting based on individual pod images. This paper attempted to develop a method that would combine the density estimation-based methods and the convolution neural network (CNN)-based methods to accurately estimate the seed count from an individual soybean pod image with a single perspective. In this paper, we first introduced a new large-scale seed counting dataset, named Soybean-pod. The dataset contains 500 annotated pod images with a total of 32 126 seeds and is the largest annotated dataset for soybean seed counting so far. Simultaneously, we used annotation information to generate a ground-truth density map by convolving a Gaussian kernel and, then, devised a simple but effective method that would elucidate pod images to a seed density map using a two-column CNN (TCNN) and thus accomplish seed counting ultimately. We conducted relevant experiments from three aspects on the new dataset to verify the effectiveness of our model and method, which provided 13.21 mean absolute error (MAE) and 17.62 mean squared error (mse). In addition, our research results showed that deep learning techniques can be easily adapted to precision tasks for plant phenotyping and breeding purposes.

Background: Effective soybean seed phenotyping demands large-scale accurate quantities of morphological parameters. The traditional manual acquisition of soybean seed morphological phenotype information is error-prone, and time-consuming, which is not feasible for large-scale collection. The segmentation of individual soybean seed is the prerequisite step for obtaining phenotypic traits such as seed length and seed width. Nevertheless, traditional image-based methods for obtaining high-throughput soybean seed phenotype are not robust and practical. Although deep learning-based algorithms can achieve accurate training and strong generalization capabilities, it requires a large amount of ground truth data which is often the limitation step. Results: We showed a novel synthetic image generation and augmentation method based on domain randomization. We synthesized a plenty of labeled image dataset automatedly by our method to train instance segmentation network for high throughput soybean seeds segmentation. It can pronouncedly decrease the cost of manual annotation and facilitate the preparation of training dataset. And the convolutional neural network can be purely trained by our synthetic image dataset to achieve a good performance. In the process of training Mask R-CNN, we proposed a transfer learning method which can reduce the computing costs significantly by finetuning the pre-trained model weights. We demonstrated the robustness and generalization ability of our method by analyzing the result of synthetic test datasets with different resolution and the real-world soybean seeds test dataset. Conclusion: The experimental results show that the proposed method realized the effective segmentation of individual soybean seed and the efficient calculation of the morphological parameters of each seed and it is practical to use this approach for high-throughput objects instance segmentation and high-throughput seeds phenotyping.

Effective and efficient fruit detection is considered crucial for designing automated robot (AuRo) for yield estimation, disease control, harvesting, sorting, and grading. Several fruit detection schemes for designing AuRo have been developed during the last decades. However, conventional fruit detection methods are deficient in the real-time response, accuracy, and extensibility. This paper proposes an improved multi-task cascaded convolutional network-based intelligent fruit detection method. This method has the capability to make the AuRo work in real time with high accuracy. Moreover, based on the relationship between the diversity samples of the dataset and the parameters of neural networks' evolution, this paper presents an improved augmented method, a procedure that is based on image fusion to improve the detector performance. The experiment results demonstrated that the proposed detector performed immaculately both in terms of accuracy and time-cost. Furthermore, the extensive experiment also demonstrated that the proposed technique has the capacity and good portability to work with other akin objects conveniently.