In recent years, new professions like "melon-tapping experts" have emerged in droves. Zhang Linghui from Handan, Hebei Province, is one such expert, whose job is to select qualified watermelons using her eyes and hands. Zhang Linghui explains that a loud sound when tapping a watermelon indicates it is ripe, while the sound of an unripe melon is relatively crisp. "Every melon has a different sound." Unripe melons have hard, inelastic rinds that make your fingers hurt when tapped; overripe melons produce a dull sound and often have cracks in the flesh when cut open, which will spoil after two days and fail to meet sales requirements. Recently, Panggezhuang in Beijing has entered the peak season for watermelon sales. When orders are large, "melon-tapping experts" like Zhang Linghui need to tap tens of thousands of watermelons every day. As early as July 2022, CCTV Finance reported that "melon-tapping experts earn 30,000 yuan a month, picking a watermelon in 3 seconds", which sparked heated discussions. Judging melons by sound is actually a scientific method To test the ripeness of a watermelon without damaging it, the tapping method is actually the most scientific. Its testing principle is that the propagation of sound waves in a medium is affected by the density and texture of the medium, so the relationship between the sound waves obtained by tapping the watermelon and the quality of the watermelon can be found. This allows testing its ripeness and internal quality without cutting the watermelon open. The internal structure of a watermelon is divided into the rind and the flesh. Structurally, the mechanical properties of different parts of the watermelon originating from different regions vary greatly. During the ripening process of a watermelon, the hardness and elastic modulus (what we commonly know as elasticity) of the rind improve, making the rind tougher. Inside the flesh, under the action of cellulase, the cellulose in the watermelon is degraded. Some cells even break away from the network woven by cellulose and become scattered, resulting in the elastic modulus of the flesh becoming smaller and smaller. To get a quantitative and analyzable conclusion, we can assume that a watermelon is a sphere in a vacuum, with a structure of a multi-layered spherical elastic body. When considering the tensile and compressive deformation of a tapped watermelon, due to axial symmetry, we only need to consider the vibration equation in one direction. If we don't need to pursue specific coefficients, we can even assume the entire watermelon as a rod. Using Newtonian mechanics and material mechanics, we can derive the vibration equation of the rod: Where U is the transverse displacement, E is Young's modulus, I is the moment of inertia about the axis perpendicular to the rod and passing through the centroid of the cross-section, and λ is the mass per unit length. Those who want to know more can search for the experiment of measuring Young's modulus by the dynamic method. The same string has many vibration modes. By expanding the vibration into different modes to solve the vibration equation, through a series of not too complex calculations such as the method of separation of variables, we can get: So here comes the key! We have obtained the formula for watermelon ripeness! For melons of the same weight, we just need to tap them and listen to the sound to know which one is riper! Of course, researchers have long tried this set of theories and conducted experiments to set a standard: A vibration frequency greater than 189Hz indicates an "unripe" melon; A vibration frequency between 160Hz and 189Hz indicates a "properly ripe" melon; A vibration frequency between 133Hz and 160Hz indicates a "ripe" melon; A vibration frequency less than 133Hz indicates an "overripe" melon. In terms of auditory perception, the higher the Hertz, the sharper the sound, and vice versa, the lower the sound. It can thus be deduced that when tapping a watermelon, a crisper sound indicates an unripe melon, and a relatively dull sound indicates a ripe melon. Therefore, if you tap a watermelon gently and it makes a crisp sound, it may be unripe; if it makes a muffled, heavy "boom-boom" sound, it should be ripe; and if it makes a "pu-pu" sound, it is overripe. Using "light" for more efficient watermelon testing In addition to manual tapping to "test melons", in recent years, some non-destructive testing technologies that can be used for large-scale and rapid testing of watermelons have gradually emerged, helping China's watermelon industry move further towards mechanization and intelligence. At the 12th Spectroscopy Network Conference held in mid-June, Researcher Huang Wenqian from the Agricultural Intelligent Equipment Research Center of the Beijing Academy of Agriculture and Forestry Sciences introduced an efficient watermelon detection technology based on full-transmission near-infrared spectroscopy. With the help of a series of technologies, watermelons can quickly complete quality testing such as sugar content, ripeness, and whether they are hollow while passing through the equipment. In the demonstration video, watermelons are placed on the conveyor belt base, passing through the detection equipment quickly and smoothly, with each melon completing quality inspection in almost a few seconds. According to Huang Wenqian, light can enter the inside of the tested object and carry out useful information. Based on this physical basis, they independently developed an online visible/near-infrared spectroscopy system called OnlineNIR with full-transmission multi-points. This system can obtain information about hydrogen-containing groups in the sample by irradiating the sample with a special incandescent lamp. With the support of this information, a chemometric model is established to realize non-destructive testing of the sample's sugar content, acidity, and internal defects. Is non-destructive watermelon testing reliable? The so-called non-destructive testing refers to testing without damaging the fruit. When the ripeness of a fruit changes, or when it is damaged or diseased, the acoustic characteristic parameters will change. Using the sound signals generated by tapping fruits to judge their ripeness, such as apples, pears, mangoes, etc., there has been a certain research foundation in the academic community. Previously, the sound-listening method was one of the traditional methods to judge whether a watermelon is ripe. The harvesting of watermelons mainly relied on the experience of farmers: tapping the rind with hands, if the sound is crisp, it is an unripe melon; if the sound is dull, it is a ripe or overripe melon. However, this method relies on farmers' experience, with low accuracy and poor reliability. Mao Jianhua developed a set of acoustic detection devices that can tap watermelons, collect sound signals, and transmit them to a computer for analysis. In this way, the farmer's ears are replaced by microphones, the hands by tapping balls, and the brain for decision-making and control is replaced by corresponding hardware circuits and analysis software, making the whole process independent of human experience. Compared with detection methods such as laser and nuclear magnetic resonance, researchers believe that the LS-SVM method based on acoustic characteristic detection is low in price and high in accuracy. Watermelons with different ripeness have different firmness (ripe watermelons have softer flesh), and thus produce different sounds when tapped. In the experiment, Mao Jianhua used 190 samples of Kirin watermelons and established five prediction models for watermelon firmness based on parameters such as the quality of the watermelon, resonance frequency, and frequency centroid within a specified frequency band. One of them is a genetic neural network model using artificial intelligence algorithms. The neural network used in this model has three layers. Mao Jianhua used acoustic parameters as input and the slope of the watermelon as the target output. After training, the prediction model for watermelon firmness was obtained. He pointed out in his paper that this model only used a single factor for modeling and prediction, and the prediction effect was closely related to the selection of initial parameters. Although the overall performance of the model is slightly lower than the stepwise linear regression model and the principal component regression model, it is far higher than the other two models, showing certain advantages. At the same time, Mao Jianhua also applied four machine learning and pattern recognition algorithms to the classification of watermelon ripeness and the identification of hollow watermelons. The sound characteristic parameters in ripeness classification are the frequency centroid within a specified frequency band and the centroid of the square of the frequency within a specified frequency band. For the identification of hollow watermelons, in addition to the above two parameters, the energy ratio is also added. The results show that an algorithm called LS-SVM can better distinguish unripe, ripe, and overripe melons, with a classification accuracy of over 70% in both the modeling set and the prediction set samples. This algorithm also performs well in identifying hollow melons, with an identification accuracy of over 90% in both the modeling set and the prediction set. Sources: NDT (Non-Destructive Testing), ViaX Salt Interest