Blackberry Fruit Classification in Underexposed Images Combining Deep Learning and Image Fusion Methods.

Berry production is increasing worldwide each year; however, high production leads to labor shortages and an increase in wasted fruit during harvest seasons. This problem opened new research opportunities in computer vision as one main challenge to address is the uncontrolled light conditions in greenhouses and open fields. The high light variations between zones can lead to underexposure of the regions of interest, making it difficult to classify between vegetation, ripe, and unripe blackberries due to their black color. Therefore, the aim of this work is to automate the process of classifying the ripeness stages of blackberries in normal and low-light conditions by exploring the use of image fusion methods to improve the quality of the input image before the inference process. The proposed algorithm adds information from three sources: visible, an improved version of the visible, and a sensor that captures images in the near-infrared spectra, obtaining a mean F1 score of 0.909±0.074 and 0.962±0.028 in underexposed images, without and with model fine-tuning, respectively, which in some cases is an increase of up to 12% in the classification rates. Furthermore, the analysis of the fusion metrics showed that the method could be used in outdoor images to enhance their quality; the weighted fusion helps to improve only underexposed vegetation, improving the contrast of objects in the image without significant changes in saturation and colorfulness.

Adaptive Feature Extraction for Blood Vessel Segmentation and Contrast Recalculation in Laser Speckle Contrast Imaging.

Microvasculature analysis in biomedical images is essential in the medical area to evaluate diseases by extracting properties of blood vessels, such as relative blood flow or morphological measurements such as diameter. Given the advantages of Laser Speckle Contrast Imaging (LSCI), several studies have aimed to reduce inherent noise to distinguish between tissue and blood vessels at higher depths. These studies have shown that computing Contrast Images (CIs) with Analysis Windows (AWs) larger than standard sizes obtains better statistical estimators. The main issue is that larger samples combine pixels of microvasculature with tissue regions, reducing the spatial resolution of the CI. This work proposes using adaptive AWs of variable size and shape to calculate the features required to train a segmentation model that discriminates between blood vessels and tissue in LSCI. The obtained results show that it is possible to improve segmentation rates of blood vessels up to 45% in high depths (≈900 µm) by extracting features adaptively. The main contribution of this work is the experimentation with LSCI images under different depths and exposure times through adaptive processing methods, furthering the understanding the performance of the different approaches under these conditions. Results also suggest that it is possible to train a segmentation model to discriminate between pixels belonging to blood vessels and those belonging to tissue. Therefore, an adaptive feature extraction method may improve the quality of the features and thus increase the classification rates of blood vessels in LSCI.