In view of the current absence of any deep learning algorithm for shellfish identification in real contexts, an improved Faster R-CNN-based detection algorithm is proposed in this paper. It achieves multiobject recognition and localization through a second-order detection network and replaces the original feature extraction module with DenseNet, which can fuse multilevel feature information, increase network depth, and avoid the disappearance of network gradients. Meanwhile, the proposal merging strategy is improved with Soft-NMS, where an attenuation function is designed to replace the conventional NMS algorithm, thereby avoiding missed detection of adjacent or overlapping objects and enhancing the network detection accuracy under multiple objects. By constructing a real contexts shellfish dataset and conducting experimental tests on a vision recognition seafood sorting robot production line, we were able to detect the features of shellfish in different scenarios, and the detection accuracy was improved by nearly 4% compared to the original detection model, achieving a better detection accuracy. This provides favorable technical support for future quality sorting of seafood using the improved Faster R-CNN-based approach.

Three kinds of nozzles normally used in industrial production are numerically simulated, and the structure of nozzle with the best jetting performance out of the three nozzles is optimized. TheR90 nozzle displays the most optimal jetting properties, including the smooth transition of the nozzle’s inner surface. Simulation results of all sample nozzles in this study show that the helix nozzle ultimately displays the best jetting performance. Jetting velocity magnitude alongYandZcoordinates is not symmetrical for the helix nozzle. Compared to simply changing the jetting angle, revolving the jet issued from the helix nozzle creates a grinding wheel on the cleaning surface, which makes not only an impact effect but also a shearing action on the cleaning object. This particular shearing action improves the cleaning process overall and forms a wider, effective cleaning range, thus obtaining a broader jet width.

The effects of different heat treatments on abalones’ texture properties and sensory characteristics were studied. Thermal processing of abalone muscle was analyzed to determine the optimal heat treatment condition based on fuzzy evaluation. The results showed that heat treatment at 85°C for 1 hour had certain desirable effects on the properties of the abalone meat. Specifically, a back propagation (BP) neural network was introduced to predict the equations of statistically significant sensory hardness, springiness, and smell using the texture data gained through TPA (texture profile analysis) experiments as input and sensory evaluation data as the desired output. The final outcome was that the predictability was proved to be satisfactory, with an average error of 6.93%.

The study of hyperspectral imaging in tandem with spectral preprocessing and neural network techniques was conducted to realize Colla Corii Asini (CCA, E’jiao) adulteration discrimination. CCA was adulterated with pig skin gelatin (PSG) in the range of 5–95% (w/w) at 5% increments. Three methods were used to pretreat the original spectra, which are multiplicative scatter correction (MSC), Savitzky-Golay (SG) smoothing, and the combination of MSC and SG (MSC-SG). SPA was employed to select the characteristic wavelengths (CWs) to reduce the high dimension. Colour and texture features of CWs were extracted as input of prediction model. Two kinds of artificial neural network (ANN) with three spectral preprocessing methods were applied to establish the prediction models. The prediction model of generalized regression neural network (GRNN) in tandem with the MSC-SG preprocessed method presented satisfactory performance with the correct classification rate value of 92.5%. The results illustrated that the integration of preprocessing methods, hyperspectral imaging features, and ANN modeling had a great potential and feasibility for CCA adulteration discrimination.

On-line measuring device of cylindricity error is designed based on two-point method error separation technique (EST), which can separate spindle rotation error from measuring error. According to the principle of measuring device, the mathematical model of the minimum zone method for cylindricity error evaluating is established. Optimized parameters of objective function decrease to four from six by assuming thatcis equal to zero andhis equal to one. Initial values of optimized parameters are obtained from least square method and final values are acquired by the genetic algorithm. The ideal axis of cylinder is fitted in MATLAB. Compared to the error results of the least square method, the minimum circumscribed cylinder method, and the maximum inscribed cylinder method, the error result of the minimum zone method conforms to the theory of error evaluation. The results indicate that the method can meet the requirement of engine cylinder bore cylindricity error measuring and evaluating.

In this study, changes in the apparent characteristics of abalone under a range of boiling temperatures and times were assessed. The trends of shape, color, and texture were statistically analyzed, while the Back Propagation (BP) neural network model was established by monitoring these 3 characteristics under different times and temperatures. This achieved a model of the characteristic parameters to predict the optimum boiling time and temperature, which can be used as a reference for abalone-processing technology. The results show that, although the model is acceptable, the BP neural network model with color feature offered the best predictor rate at 81.74%.