Aiming at the inherent limitations of traditional single-point spectral detection methods in terms of simultaneously achieving population target localization and multi-dimensional quality characterization, as well as the significant engineering challenge where the dynamic extraction of spectral Region of Interest (ROI) is highly susceptible to background noise interference in unstructured industrial environments, a comprehensive multi-dimensional quality detection and commodity grading framework (abbreviated as YOLO-HSI) based on the synergistic perception of an oriented object detection network (YOLO-OBB) and push-broom hyperspectral imaging (HSI) was proposed for Kashi prunes to achieve rapid non-destructive inspection.By means of introducing the deep-learning-based YOLO-OBB network combined with a newly developed geometric homomorphic contraction strategy, this study successfully overcame the technical bottlenecks concerning high-precision spatial localization for fruits with arbitrary random tracking postures and the dynamic extraction of pure, unpolluted spectral profiles. On the basis of this spatial-spectral synergy, a comprehensive evaluation system encompassing external dimension measurement, internal soluble solids content (SSC) quantitative inversion, and early latent moldy defect elimination was effectively constructed. To implement this proposed quality assessment framework, consecutive data optimization involving Savitzky-Golay smoothing, Min-Max normalization, and second-derivative transformation was rigorously conducted on the extracted pure ROI spectra, followed by feature variables selection and the formulation of a multi-indicator comprehensive grading rule according to national fruit quality guidelines. The comprehensive experimental results demonstrated that the optimized YOLOv11n-OBB model achieved outstanding localization and detection capabilities, yielding high detection recall rates of 0.992 for intact prunes and 0.982 for moldy prunes, while achieving ultra-fast online processing speeds of 243 frames per second and 270 frames per second, respectively, which reflects a significant advantage for real-time high-throughput inspection. Based on the pixel-to-physical world coordinate mapping relationship, automated millimeter-level measurements of fruit polar diameter and equatorial diameter were successfully realized, where the size prediction root mean square errors (RMSE) fell below 1.1 mm for intact prunes and 0.8 mm for moldy prunes. In terms of internal quality detection, the classification model based on second-derivative preprocessing achieved a complete discrimination accuracy of maximum and an ideal area under the curve (AUC) score of 1 for identifying early latent moldy fruits, while successfully compressing spectral variables to achieve a feature wavelength reduction rate of 70.1%. Furthermore, the CARS-Ridge multivariate regression model exhibited robust predictive performance for internal soluble solids content, yielding a high prediction coefficient of determination (R²_p) of 0.7225 and a reliable ratio of performance to deviation (RPD) of 1.90. The multi-indicator grading decision framework demonstrated that the comprehensive commodity grade determination accuracy reached 85.00% for premium, first-grade, and second-grade prunes, with cross-grade misclassifications occurring exclusively between adjacent boundary levels without severe grade-skipping deviations. In conclusion, the findings of this study successfully validate the high effectiveness and feasibility of the image-spectrum synergistic technical route for the non-destructive detection of multi-dimensional quality in Kashi prunes. The excellent detection performance and classification accuracy achieved by the proposed framework can provide core algorithms and data-driven theoretical support for the research, development, and industrial edge deployment of a new generation of intelligent fruit sorting equipment.

Cotton is one of the most crucial global textile raw materials to determine the quality of final products. The moisture content of seed cotton can profoundly impact the storage, transportation, and textile processing of cotton. This study aims to rapidly and non-destructively measure the moisture content of seed cotton. A quantitative detection model was established using Fourier-transformed near-infrared spectroscopy. A series of experiments were conducted to explore the influence of seed cotton sample density on spectral curves. Sample density was significantly dominated in the spectral curves, where the lower densities resulted in stronger spectral signals. However, the fluctuations were stabilized in the spectral curve, when the sample density reached a specific threshold 0.088 60 g/cm³. Data accuracy and comparability were the pivotal reference points. Subsequently, Fourier-transformed near-infrared spectroscopy was employed to collect the absorbance spectral data of seed cotton samples within the range of 3 900-11 000 cm⁻¹ wavelength. The samples were dried in an air blast drying oven (DWG-9240A) under constant temperature. The airflow was used to remove the moisture from the seed cotton. Pre-experimental verification showed that the weight of samples remained relatively unchanged, when the temperature was raised to (105 ±3) ℃ and maintained for three hours. Afterward, the dried samples were removed and weighed to measure their moisture content using a balance with a precision of 0.001 g. Furthermore, nine preprocessing methods were applied to the original spectral data, in order to enhance the data quality. Comparative analysis determined that the best performance was achieved in the first-order derivative combined with detrending (FD-DT) preprocessing using the Partial Least Squares Regression (PLSR) model. The calibration and prediction set determination coefficients were 0.974, and 0.845, respectively, with the root mean square errors of 0.316 and 0.721, respectively, as well as the residual prediction deviation (R_PD) of 3.00 after FD-DT preprocessing. The wavenumber range of 4000-10000 cm⁻¹ was selected to extract feature spectral data. The reason was that the lower sensitivity and response of the spectrometer at the beginning and end of the spectra, led to weaker or unstable signals in these regions. The optimal feature wavelengths were then obtained using Competitive Adaptive Reweighted Sampling (CARS), Information Gain (IG), Successive Projections Algorithm (SPA), and Pearson's Correlation Coefficient (CC). The feature wavelength counts of 47, 27, 30, and 27, accounting for 6.03%, 3.46%, 3.85%, and 3.46% of the spectral range, respectively. These features effectively reduced the number of variables for the high efficiency and performance of the model. After the extraction of feature wavelength, the quantitative models were established to predict the moisture content of seed cotton using Partial Least Squares Regression (PLSR) and Support Vector Machine (SVM). Comprehensive analysis of various analytical algorithms showed that the combination of FD-DT-CARS-PLSR and FD-DT-CARS-SVM was the most effective predictive model, where the determination coefficients of 0.933 and 0.931, root mean square errors of 0.480 and 0.500, and residual prediction deviations of 3.88 and 3.85 in the prediction dataset. FD-DT was used to effectively remove the trends and noise from the data for data quality and usability. CARS was used to efficiently select the most relevant feature wavelengths for the performance and prediction accuracy of the model. PLSR demonstrated excellent performance on multicollinear data for better interpretability, but with a relatively weaker performance in fitting nonlinear data. SVM displayed strong capabilities in nonlinear modeling and adaptability to high-dimensional data, but it was relatively difficult to interpret with the slower training times for large datasets. Two models can be combined to effectively predict in this case. In summary, near-infrared spectroscopy can be expected to rapidly, non-destructively, and accurately detect the moisture content of seed cotton .

This study aims to reduce the violent vibration for the high reliability of the integrated machine for straw return and residual film recycling. A vibration characteristic analysis was carried out to optimize the structural parameters. The modal characteristics of the frame were solved using the Lanczos algorithm on the ANSYS software, in order to obtain the modal frequencies and shapes. The equivalent vibration mechanics model was used to establish the vibration model of the machine. Newton's second law was adopted to analyze the model force. The vibration program of the simulation was constructed to solve the vibration parameters using Matlab/Simulink. A series of measurements of point locations were used to verify the vibration of the machine. The test was also carried out in the conditions of no-load and field harvesting. Some sensors were installed and connected to the YND Data Acquisition Instrument. Vibration data was collected with the time and frequency domains. The results showed that the 1st and 2nd order natural frequencies of the frame were 39.001 and 39.076 Hz, respectively. The modal vibration pattern was that the side panels oscillated back and forth along the y-axis. The vibration intensity under field harvesting was higher than that under no-load conditions. The measurement point 3 vibration amplitudes in the two conditions were 34.93 and 22.36 m/s², respectively, where the difference in the vibration amplitude was 12.57 m/s². The maximum was found in the measurement point 3 of the Y, and Z directions in field harvesting, as well as the measurement point 5 in the X direction of the amplitude. The continuous operation of the stripping device increased the vibration intensity of the machine with less reliability in the field harvesting under the straw crushing device. The straw-crushing device and the stripping device were the main components that contributed to the machine's vibration. The effective values of vibration acceleration were 2.92 and 2.64 m/s², respectively, after test and simulation. The relative error of 9.6% fully met the requirements of mechanical simulation. The spectrum analysis showed that the machine vibration frequency was mainly for the fundamental frequency of the tractor power output shaft, as well as the fundamental frequency and frequency multiplier of the straw crushing device. The large vibration intensity was measured at point 3 of the X, Z directions, and point 1 of the Z direction. There were weld cracks and bolt loose. The measured vibration frequency appeared at 36.13 Hz close to the first two orders of the frame's natural frequency. The modal vibration pattern was easy to stimulate the local resonance in the range of excitation frequency. The orthogonal experiment was adopted to optimize the frame with the thickness of the side plate, the thickness of the main connecting beam tube wall, and the width of the secondary support beam as the object variables, while the 1st and 2nd orders of the frame's natural frequency as the indicators. The optimal parameters were achieved, where the thickness of the side plates was 12.0 mm, the tube wall thickness of the main connecting beam was 6 mm, and the width of the secondary support beam was 70 mm. The 1st and 2nd natural frequencies were 50.700 and 53.322 Hz, respectively, which were higher than the maximum limit of the excitation frequency in the straw-crushing device of 46 Hz. Thus the resonance was effectively avoided. The improved model was verified to significantly reduce the vibration intensity of each measurement point, with a 48% amplitude of the maximum reduction. The transmission of vibration intensity was also significantly weakened, indicating the effective optimization. This finding can provide the theoretical reference for the vibration analysis and structural optimization of the residual film recycling machine.