Siti Khairunniza-Bejo

Coffee, specifically Coffea arabica, is highly valued for its quality, aroma, and flavor. Understanding the factors affecting coffee quality is essential for farmers, industry players, and consumers. One key factor is the storage conditions, including air exposure, light, moisture, and temperature. Moisture content (MC) and color are commonly used indicators of coffee quality. This study analysed MC and colour trends of dark and medium roasted Arabica ground coffee stored under various conditions (cold vs. room temperature, airtight vs. non-airtight, and opaque vs. transparent) over a 10-week period. The results showed that the MC increased under all conditions. For dark roast, the lowest MC was observed in cold, airtight, transparent storage, whereas the highest was observed in room-temperature, non-airtight, opaque storage. For medium roast, the lowest MC occurred in cold, airtight, opaque storage, and the highest occurred in room-temperature, non-airtight, transparent storage. Color analysis using RGB imaging provided reliable grayscale values consistent with visual observations, although the trends fluctuated and were not strictly increasing or decreasing. Nevertheless, the method effectively distinguished roast levels, with dark roast showing lower grayscale values than the medium roast. A relationship between MC and color was also observed, where a higher MC corresponded to higher grayscale values (slightly brighter appearance). Overall, this study highlights the impact of storage conditions on MC and color, offering insights for improving coffee storage practices to maintain quality and freshness.

Mangroves play a critical role in global carbon sequestration, biodiversity conservation, and climate change mitigation. Accurately quantifying mangrove biomass is essential for sustainable forest management and carbon accounting. Yet, the structural complexity and species diversity of mangrove ecosystems pose significant challenges for accurate estimation. In this study, we developed an integrated model that combines multispectral imagery and radar data. Using Sentinel-2 and ALOS-2 satellite imagery combined with field measurements, these data were used to construct linear regression and random forest models for the Matang Mangrove Reserve, Malaysia. We further analyzed the relationships between vegetation indices, radar polarization modes, and biomass. Results indicate that the average biomass is approximately 146 t/ha. The Optimized Soil-Adjusted Vegetation Index (OSAVI) and horizontal–vertical (HV) polarization showed the strongest correlation with field-measured biomass, with an R² of 0.735 and a root mean square error (RMSE) of 46.794 t/ha. This study provides a scientific basis and technical support for mangrove carbon stock assessment, ecosystem management, and climate change mitigation strategies, and highlights the potential of integrating optical and radar remote sensing for large-scale mangrove biomass monitoring.

The weedy rice contamination in certified rice seed samples greatly impacts the Malaysian rice seed industry. The existing manual process to identify the weedy rice seed is only based on the physical appearance of the seed. The physical characteristics such as morphology, colour and texture image of the seeds were captured and analysed using image processing and the application of machine vision to understand the physical characteristics of the weedy rice seed. The objective of this study is to understand the physical characteristics of the weedy rice seed using Principal Component Analysis (PCA) transformation. A total of 7350 images of cultivated rice seeds from five major varieties and 895 images of weedy rice seeds were acquired using machine vision setup, and 67 features from the three major parameters (morphology, colour, and texture) were extracted. The test of equality of means based on Wilks' Lambda was performed to assess significant differences among the group parameters. PCA transformed data into principal components. The relationships between weedy rice seed and cultivated rice seed samples were examined through the score and loading plots of the PCA analysis. This newly transformed data visualises the experimental data's underlying structure and helps identify the parameters distinguishing between weedy rice and cultivated rice seed. The results on the PCA score plot have shown overlapping areas between the cultivated and weedy rice seeds, indicating high similarities between the seed samples.

Paulownia has gained recognition as one of the swiftest-growing tree species globally and is used for medicinal, ornamental and timber purposes. The current conventional method of Paulownia tree counting is based on the ground manual survey which is timeconsuming. Therefore, the main objective of this study was to develop a suitable model for Paulownia tree counting using template matching and unmanned aerial vehicle imageries. First, the suitable template matching method was identified by comparing the performance of template matching with georeferencing data and template matching with the predetermined size of the template image. It was then followed by the development of an automatic Paulownia tree detection and counting model using the most suitable template matching technique at 4 different growing stages which are 3, 6, 9 and 12 months old. Results have shown that the template matching with georeferencing data performed better compared with the template matching with the predetermined size of the template image. For different growing stage models, the 3 and 6-month-old Paulownia tree models perform better compared to the others with an average F1-score of more than 85%. Results also revealed that younger Paulownia present greater homogeneity in crown morphology thus making it easier for the algorithm to detect Paulownia tree at younger age.

Mangoes are highly vulnerable to various diseases, both before and after harvest. One of the fungal maladies impacting mango trees is mango anthracnose, stemming from the pathogenic agent Colletotrichum gloeosporioides. Infected mango fruit typically exhibit dark, sunken lesions and black spots on their skin, which diminishes their quality and leads to considerable losses in the mango market. Therefore, this study aimed to characterise the physicochemical properties of Milk Gold mangoes infected with anthracnose disease. The samples were grouped into two: control samples and treated samples. The treated samples were inoculated with a spore solution of C. gloeosporioides (2.62 × 106 cell/ml) and incubated at room temperature (28ᵒC) for 7 days while control samples were injected with distilled, sterilised water. The progress of the anthracnose disease was evaluated at day 0 (after 1 hr of injection), 3, 5 and 7 of incubation period, and various physicochemical parameters, such as total soluble solids content (TSS), firmness, pH, colour, weight loss, and moisture content (MC) were measured. The results indicated that the anthracnose disease manifested as small black spots after 3 days of storage, and the flesh of the fruit began to rot severely by day 5 of storage. The increasing trends were observed in the TSS, pH, weight loss, and moisture content of the fruits, concomitant with a decrease in fruit firmness as ripening advanced. Notably, all colour parameter values (L*, a*, b*, c* and H*) indicated significant differences between the inoculated and control samples. With the rapid development of the symptoms, it is crucial to identify the disease at an early stage to apply the appropriate disease control measures during post-harvest handling.

Recently, transformer-based models have achieved remarkable performance in the hyperspectral image (HSI) classification. However, due to the limited training data, existing methods often show limited capability of capturing fine-grained local features. Although attempts have been made to solve this problem, the large amount of parameters imposes the risk of overfitting. In this paper, we propose a locally enhanced transformer network for HSI classification with fewer network parameters, which mainly consists of a multi-branch spatial-spectral tokenization (MSST) module and a dual-branch transformer encoder (DTE) module. The MSST generates effective spatialspectral tokens through diverse convolutions with a residual connection. The DTE consists of a global transformer branch and a locally enhanced transformer branch, which are used to capture the global and local spatial dependencies of HSI, respectively. Unlike the conventional self-attention module used in the global branch, we propose an improved multi-head selfattention (IMSA) module in the local branch by incorporating the local prior information of HSI with graph convolution, to enhance the local information extraction. To fuse the global and local features from the two branches, we introduce an adaptive strategy by using learnable weights for both branches. We devise our MSST and DTE with a shallow architecture, significantly reducing the number of parameters. Experimental results on benchmark datasets demonstrate that the proposed method outperforms the state-of-the-art.