Search

Industrial seedling quality assessment, such as attempting to find abnormal seedlings, is a challenging task where assessment methods must contend with the natural variability of seedlings, as well as the subjective nature of expert judgements. Furthermore, obtaining expert judgements is expensive and time-consuming, so machine learning approaches which rely on fewer judgements would be useful in practice. We investigate autoencoders, operating on 3D point clouds obtained from 6732 seedlings to address this challenge, exploiting such systems’ ability to work with partially labelled data. Point clouds from tomato seedlings are recorded using a 3D data capture platform, MARVIN™, and the quality of each seedling is determined by expert judgement. An existing system is used to establish baseline performance scores using a rule-based expert system and machine learning with handcrafted features. Autoencoders are trained on the point clouds to learn representations for subsequent use in classification. We examine scenarios where large amounts of partially labelled data are available, and compare with the case where fully labelled data is available. To improve performance, we compare the architectural subcomponents based on PointNet and PointNet++, as well as the effect of different training strategies. We find, with 13.6% of training data labelled, our model has correct classification rates of 97.7% and 82.7% for normals and abnormals respectively. With further improvements and fully labelled data, we find that correct classification rates of 97.6% and 96.1% can be reached. The results demonstrate that semi-supervised learning supported by partially labelled data has the potential to greatly reduce the cost of data curation, with minimal impact on overall accuracy.

Seaweed is increasingly becoming a crop of interest in aquaculture. Seaweed has potential as a low trophic food source as no fresh water or fertiliser is needed for its growth. However, to start a profitable business in seaweed farming in the North Sea, space is required. Besides that, offshore seaweed farming is met with numerous technical challenges in this sometimes-turbulent environment. To monitor the crop farmers can make use of the latest technology in remote sensing. Potential remote sensing technologies, which could be used in seaweed farming, were identified in 2020. The use of the DIDSON sonar was thought in advance to make the most useful underwater images of the seaweed Saccharina latissima in a test farm in the Oosterschelde. Were in 2020 the first preliminary images were made with the DIDSON. In 2021 the DIDSON was used to make images in the seaweed test farm, in combination with using Humminbird sonar (Helix 12, MSI/GPS G3N). In situ, it was found that handling the DIDSON while in a small boat was a difficult task. Because of movement of the sonar, it was difficult to make sharp images. The DIDSON again did not yield a lot of useful images. However fully convolutional neural network models for image recognition were tested using images from both years. The Humminbird fish finder was not successful in taking any images of seaweed in 2021, though the mussel lines in the same farm could be detected. In 2022 the last sampling was done again using both sonars. This time using different settings, the Humminbird was able to detect the seaweed in the lines in the farm. The images that the Humminbird yielded had a better resolution and quality than those of the DIDSON from the previous years. This data could be used on further expanded neural networks. The images of the Humminbird were used in a classical approach for segmenting in the neural network and showed promise for future use. However the data had a lot of limitation and in follow up studies multiple

Bacterial and virus diseases causes major damage in agriculture. Classifying those is an challenging but important task. This dataset contains +-274 000 spectral line images (512 pixels x 56 bands) of 6 different cultivars of healthy, bacterial (Erwinia) and virus (PVY) infected plants. The coordinates of each image is stored and can be compared with the stored position of individual plants labeled by crop experts