Citrus rich in nutrients is one of the most favorite fruits in recent years. But the thick-skinned citrus is often accompanied by the fruit hollow, skin thick floating, leading to the low actual volumetric edible rate. Furthermore, traditional nondestructive testing cannot accurately and rapidly detect the volumetric edible rate. In this study, the line array X-ray image acquisition and three-dimensional reconstruction were developed to detect the volumetric edible rate of thick-skinned citrus fruits, including the fruit rotation and lifting, data acquisition, radiation protection and motion control. The fruit rotating device consisted of a rotating table and a gear motor; the fruit lifting device consisted of a lifting platform and a stepping motor. The data acquisition device included an X-ray emission source, a line array X-ray detector, and upper computer software. The radiation protection device was a lead plate with a thickness of 2mm, which was used to prevent the radiation generated by X-rays from leaking to the external environment. The motion control part consisted of a programmable logic controller and upper computer software. Taking the Ugli fruit as the detection object, the information entropy of the X-ray projection map was evaluated to optimize the detection parameters. The optimal tube voltage and current of X-ray source were obtained to be 67 kV and 0.92 mA, respectively. The integration time of the line array detector was 1 ms. Dark-field correction and bright-field compensation were used to remove the uneven pixel distribution and the noisy background in the initial state of the X-ray detector before the experiment. A series of 180 X-ray projections were collected in the circumferential direction at intervals of 2.0° rotation. The samples were rotated by 120° with the mid-axis of the fruit as the center of rotation. A series of citrus X-ray projection maps were captured at three angles of 0°, 120° and 240°, respectively. The X-ray projection maps were converted to sinograms using the Radon transform. After that, the sinograms under the three angles were reconstructed as the slice maps using the FBP (Filtered Back Projection). Image segmentation was carried out on the slice maps using image filtering, enhancement, and thresholding segmentation, in order to form the background, pericarp, pulp, and cavity region. Moreover, the regional area ratio was defined using the ratio of citrus pulp region to fruit region, whereas, the slice map edible rate was defined using the regional area ratio. The physical parameters were measured, such as transversal diameter, vertical diameter, mass, volume, density and fruit shape index of citrus fruits. The true volumetric edible rate of citrus was calculated using the specific gravity of citrus pulp to the volume of the whole fruit. There was the better correlation between the physical parameters and volumetric edible rate of citrus. The results showed that there was the higher correlation between fruit density and slice map edible rate and citrus volumetric edible rate, with the highest correlation 0.93 for slice map edible rate. Finally, the slice map edible rate was selected as the input feature of the model. The linear regression model was used to quantify the volumetric edible rate of thick-skinned citrus, with the values of $R_{\mathrm{p}}^2$ (coefficient of determination of prediction), RMSEP (root mean square error of prediction), and RPD (residual predictive deviation) of 0.86, 4.81%, and 2.71, respectively. In conclusion, it is feasible to quantitatively analyze the volumetric edible rate of thick-skinned citrus using X-ray three-dimensional reconstruction. The developed approach can also be applied in the nondestructive testing of the quality of agricultural products. Therefore, the nondestructive testing techniques can be expected to evaluate the internal tissue lesions and external quality of agricultural products.

Salt content is one of the most important indexes to evaluate the quality of salted duck eggs. The variety, source, size, and freshness of duck eggs can significantly affect the salting rate of duck eggs. Rapid salting technology (such as high-pressure pulsation) can promote the efficiency of osmotic dehydration. The difference in the raw materials can also be magnified to result in the uneven salt content of products. Therefore, it is crucial to precisely measure the salt content for the quality consistency of salted duck eggs. Rapid non-destructive testing of poultry eggs can include machine vision, spectral, and acoustic technology at present. Among them, near-infrared spectroscopy has been employed to measure the salt content of salted eggs. But its equipment cost is higher than the others. By contrast, machine vision is normally used to differentiate between deteriorated and high-quality eggs, due to its versatility, affordability, and large-scale detection. In this study, the acquisition device of the image was proposed to accurately detect the salt content of salted duck eggs using an industrial camera and transmission light source. There was a notable improvement in the light transmittance of salted duck eggs, as the salting time increased. Additionally, the visual field area with the yolk was gradually diminished in the transmission image. The color was lightened gradually with the regular changes until it nearly disappeared. There was a gradual increase in the average egg white, egg yolk, whole egg salt content, and egg yolk index, as the salting time increased. However, there was a significant variation in the salt content and yolk index of salted duck eggs under the same salting time. The eggshell and eggshell membrane served as the primary channels and barriers for the material exchange between the duck egg and the external salt solution. The high-pressure pulsation greatly contributed to the differences in the eggshell and eggshell membrane among individual duck eggs, further impacting the osmotic dehydration rate. Therefore, the salt content of salted duck eggs was required under conditions of high-pressure pulsating salting. The prediction model was established for the salt content and yolk index of egg white, yolk, and whole egg. The overall image features and long-axis cross-section light intensity were extracted using multiple linear regression and support vector machine. Results indicated that the overall image features were used to perform better on the prediction of egg white, yolk, and whole egg salt content. While the features of long-axis cross-section light intensity were more effective for the prediction of yolk index. Notably, the optimal combination was achieved in the support vector machine with the overall image features. The salt content of egg yolk was accurately detected in the test set with the R_p, RMSEp, and RPD of 0.846 0, 0.341 6, and 1.898, respectively. Furthermore, the optimal detection of egg yolk index was obtained in the multiple linear regression model with the long-axis cross-section light intensity in the test set with R_p, RMSEp, and RPD of 0.831 8, 0.074 3, and 1.916, respectively. This finding can provide a theoretical foundation and technical support to rapidly detect the salt content and yolk index in salted duck eggs.

Detecting embryo viability is essential to the quality and safety of vaccine production, especially in large-scale manufacturing. Rapid and accurate detection of embryo viability can improve the production efficiency for the final quality of vaccines. Traditional machine vision detection can rely heavily on the complex algorithms of feature extraction, most of which are often designed for specific scenarios. However, the detection accuracy and stability are also sensitive to the image quality and environmental conditions, such as lighting, background, or temperature. Additionally, the applicability of traditional detection has been limited to fault tolerance in different environments, when dealing with noise or abnormal conditions. To address these challenges, this study aims to detect the vaccine embryo viability using an improved YOLOv8 model. Several innovations were incorporated to enhance efficiency, accuracy, and adaptability. A specialized system of image acquisition was developed to capture the high-quality images of embryos incubated for 10 to 11 days. The consistent dataset was obtained in the varying environmental conditions. The dataset was then expanded using geometric transformations, color adjustments, and image enhancement. As such, the robustness of the model increased to handle the diverse image conditions. In terms of model improvements, ShuffleNetV2 was used to replace the YOLOv8 backbone. Computational complexity was significantly reduced to maintain high accuracy, indicating more suitable for deployment on embedded devices where computational power was limited. The overall efficiency of the model was enhanced to support its application in large-scale industrial environments. Additionally, a dynamic snake convolutional layer was added to the neck of the YOLOv8 model. This layer was used to adaptively focus on the elongated and curved structures in embryos, in order to capture the geometric features of tubular structures. The precision of detection was improved to more accurately assess the physiological state of the embryos. Furthermore, the EIoU (Embedding Intersection over Union) loss function was introduced to more effectively detect the boundary box alignment and shape similarity, compared with the traditional IOU. EIoU improved the accuracy of boundary box positioning, while reducing the errors related to the complex shapes of embryos, thereby enhancing the reliability of the model in real-world applications. Experimental results confirmed that the superior performance of the improved YOLOv8 model was achieved to detect embryo viability. There was a precision of 99.2%, a recall of 98.2%, and a mean average precision (mAP_50-95) of 96.9%, with increases of 2, 0.3, and 1.5 percentage points, respectively, compared with the original YOLOv8 model. Additionally, the computational complexity and inference time were reduced by 60.9% and 60.5%, respectively. The improved model was highly suited for the large-scale detection of embryos. The finding can also provide an efficient, non-destructive approach for the rapid detection of the vaccine embryo viability.