Research On Freshness Detection Approach For Chinese Mitten Crab Based On Machine Olfaction

Aquatic products freshness detection is an important topic in the current issue of food quality and safety. As Chinese mitten crabs volatilize poisonous substances in the state of dying or death, it will pose a serious threat once human consumption. In the current market trade, traditional freshness detection of crabs is sensory detection which is often affected by the subjectivity. In the case of a large number of Chinese mitten crabs are bundled, the final results obtained are biased. So research on destructive, real-time and accurate Chinese mitten crab freshness detection technology is an urgent demand of the current aquatic products market development. In particular, research on freshness detection approach for Chinese mitten crabs based on machine olfaction has more important practical significance.Physical and chemical indicator TVB-N can detect the freshness of Chinese mitten crabs, but the crab meat to be used as the experiment subject. The entire experiment contains high-temperature distillation and many other complex aspects which can not meet the purpose of rapid market transaction. With the development of sensor technology and pattern recognition technology, the complete living detection of aquatic products achieved by machine olfactory system will have broad application prospects. The subject of Chinese mitten crabs freshness detection techniques based on machine olfaction has been studied, the main results are:The machine olfactory system was built combined with the purpose of complete living detection. Considering the shape and volume of Chinese mitten crabs, we designed the odor sampling platform by static headspace method.For the characteristics gas were volatiled in living state, we research on the selection of machine olfactory sensor arrays. For subsequent information collection and processing, the PC interface which can realize the function of each module of the software system were made.In current, there is no research on odor information collection in state of complete living body. In the experiment, such as headspace temperature, headspace volume, headspace time, scrubbing time, sampling time and other parameters were need to be analysised and studied for determining a reasonable experiment solution. Then, we sequentially performed mean filtering, baseline processing, removing abnormal data and other operations for smell information. Steady-state information and transient information were chosen for effective feature.Since the materials of sensors are non-linear metal-semiconductor, the obtained odor information must also include non-linear characteristic. Traditional linear dimension reduction algorithm can not extract the nonlinear characteristics, the subject proposed laplacian eigenmaps algorithms for extracting odor information characteristics, the final visualization classification results have great distinction compared to the traditional algorithms.TVB-N was used as freshness evaluation criteria. The relevance between odor responses and TVB-N were established by multiple linear regression model, it reflected that smell information were used to evaluate the freshness was feasible. The subject established prediction model for Chinese mitten crabs freshness based on BP neural network. Basically, we can consider that the model for evaluation of the freshness of Chinese mitten crabs is valid.