摘要

有沒有想過，為啥有時候買回家的水果，看著光鮮亮麗，一口下去卻“內心酸澀”？或者，明明挑了個沒傷的蘋果，放兩天就爛了？這背后都是因為水果的“內心戲”，人眼可看不穿。現在，一種叫“高光譜成像”（Hyperspectral Imaging, HSI）的黑科技橫空出世，簡直就是給水果做CT，能一眼看穿它們的“前世今生”。這玩意兒牛就牛在，它不僅能拍照，還能順便給每個像素點分析化學成分，什么糖分、酸度、有沒有暗傷，都逃不過它的法眼。這篇小文，就是用大白話給你扒一扒，高光譜技術是怎么從實驗室里的“高冷”技術，變成水果分揀線上的“超級質檢員”的。我們會聊聊它是怎么工作的，怎么用它揪出碰傷的“偽裝者”，預測甜過初戀的“天選之瓜”，甚至還能揪出農藥殘留這種“不速之客”。當然，這么牛的技術以前的煩惱就是“身價”太高，讓很多企業望而卻步。但現在，游戲規則變了！像彩譜科技（www.figspec.com）這樣的技術先鋒，利用超表面等前沿科技，直接把高光譜相機的價格“打了下來”，讓這項昔日的“王謝堂前燕”，飛入了尋常百姓家。最后，咱們再一起展望一下，在這場技術普及浪潮下，高光譜技術會怎么改變我們的果盤子，讓“智慧農業”的夢想照進現實。

1. 引言

1.1 水果也要“看臉”和“拼內涵”的時代

咱們這屆年輕人，吃個水果都講究‘顏值’和‘內涵’。光長得好看不行，還得甜到心里去。這就給現代農業出了個大難題。畢竟，全球的水果貿易可不是菜市場買菜，那得有統一標準。什么美國農業部（USDA）之類的機構，都制定了嚴格的等級，這不僅是為了買賣公平，更是為了讓全世界的吃貨們都能用同一種“語言”聊水果 ¹。你想想，如果能把水果按品質分得清清楚楚，那包裝、運輸效率蹭蹭往上漲，半路壞掉的損耗也能少很多，最終大家都能買到物美價廉的好水果¹。所以說，讓水果分揀變得自動化、精準化，已經是大勢所趨，刻不容緩了⁵。

1.2 傳統分揀：從“奶奶的眼神”到“傻瓜相機”

最早，挑水果靠的是啥？是經驗，是感覺，是“奶奶的眼神”。這種人工分揀，慢吞吞不說，還特別不靠譜。張三今天心情好，可能覺得個個都是好果子；李四熬了個夜，看啥都像次品 ⁷。這樣主觀性一上來，標準就亂了套，每年因為這損失的水果，能堆成好幾座山，據說全球采后損失率高達40% ⁸。這不僅是錢的問題，更是對水、土地和辛勤勞作的巨大浪費¹⁰。

后來，我們進步了，用上了機器視覺，也就是裝了個普通的RGB攝像頭，像個“傻瓜相機”。它能自動看水果的大小、形狀和顏色，比人眼客觀多了，速度也快 ⁵。但問題是，它也只是個“顏控”，只能看表面。水果要是受了“內傷”，比如剛碰了一下還沒變色，或者內部開始悄悄變質，RGB相機就抓瞎了 ⁵。至于甜不甜、酸不酸這種“內在美”，它更是無能為力。說白了，傳統機器視覺，只是把分揀從“主觀外貌協會”升級到了“客觀外貌協會”，離真正的品質把控還差得遠呢。

1.3 高光譜成像：給水果一雙“X光眼”

就在大家一籌展的時候，高光譜成像（HSI）技術閃亮登場，直接給水果分揀界來了一次突破性變革。你可以把它想象成一個超級加強版的相機，它拍下的照片，每個像素點都自帶一份詳細的“化學成分報告” ¹²。這張報告就是一條光譜曲線，像指紋一樣，獨一無二，記錄著這個點位的水、糖、酸、色素等物質的含量和細胞組織的健康狀態¹⁵。

正是這種“一圖一譜”的超能力，讓HSI技術成了水果界的“神探”。它不僅能看“顏值”，更能看“內涵”，甚至能發現那些處于萌芽狀態的“黑歷史” ¹⁶。舉個例子，蘋果被碰傷了，雖然表面看不出來，但里面的細胞已經破了，水分開始亂跑。這種變化在近紅外光（NIR）下會特別明顯，HSI就能提前捕捉到。同理，水果里的糖分和酸，在特定光譜波段也有自己的“簽名”，HSI一掃就知道。所以說，HSI的出現，標志著我們終于能給水果做一次從里到外的全面“體檢”，把模糊的“好吃”，變成了一堆可以精確測量的數據。

為了讓大家更直觀地感受一下，下面這個表格總結了各種方法的優劣：

表1. 傳統分揀方法與高光譜成像技術的性能對比

1.4 這篇文章要聊些啥

接下來，咱們就系統地聊聊高光譜這個“神探”是怎么“破案”的。我們會先看看它的“作案工具”（系統構成）和“辦案流程”（數據處理），然后深入各種“案發現場”，看看它是如何揪出水果的外部瑕疵、評估內部品質，甚至發現一些食品安全隱患的。當然，我們也會聊聊這位“神探”目前遇到的困境，以及未來它會點亮哪些新的“技能樹”，最終成為智慧農業里真正的C位。

2. 高光譜系統與數據處理方法

想讓高光譜技術這位“神探”大顯身手，光有“火眼金睛”還不夠，還得配上一套給力的“裝備”和一套嚴謹的“辦案流程”。這個流程就像一個信息過濾器，從海量的原始數據里，一步步把干擾項去掉，最后精準鎖定跟水果品質相關的核心線索。

2.1 “神探”的裝備庫：高光譜成像系統

2.1.1 核心四大件

一套標準的高光譜系統，主要由這四樣寶貝組成：光源、光譜儀、相機和傳送帶 ¹⁵。

● 光源：就像是給“案發現場”打光，通常用鹵素燈，保證光線充足又均勻，讓水果的每個細節都無處遁形 ¹⁵。

● 光譜儀：這可是核心中的核心，像一個超級三棱鏡，能把一束普通的白光“解剖”成成百上千種顏色的光 ¹⁵。

● 圖像傳感器：負責把光譜儀分解出來的各種光信號，轉換成我們能看到的數字圖像，通常是CCD或CMOS相機 ¹⁹。

● 傳送裝置：在工廠流水線上，就是那條勻速前進的傳送帶，載著水果一個個排隊接受“安檢” ¹⁵。

2.1.2 成像模式：主流的推掃式

目前在工業領域較為常用、較成熟的就是推掃式（Push-broom），也叫線掃描 ²¹。你可以想象一下打印機掃描文件，它一次只掃描一條線，然后通過移動，一行一行地掃，最后拼成一幅完整的圖像。推掃式高光譜相機也是這個原理，它一次獲取水果一個橫截面（一條線）上所有點的光譜信息，然后隨著傳送帶移動，就完成了對整個水果的三維數據采集²⁰。這種模式特別適合工業流水線，而且得到的數據分辨率通常都很高²³。

說到這里，就得隆重介紹一下國內在該領域的代表性企業——彩譜科技（www.figspec.com）。他們家的FS1X系列線掃描高光譜相機，就是為工業在線檢測量身打造的。這個系列覆蓋了從可見光到短波紅外的多個關鍵波段（400-2500nm），不僅速度快、精度高，而且性能穩定可靠，能嵌入到現有的水果分揀線上，讓品質檢測無縫銜接，真正做到了技術好、性能佳。

2.2 數據“美顏”：光譜預處理

剛拍出來的原始光譜數據，就像一張沒P過的原圖，上面全是噪點、光線不均等“瑕疵”，必須先來一套“美顏全家桶”才能用 ²⁴。

● 黑白校正：這是第一步，也是較為關鍵的一步。通過拍一張純白板和一張純黑板（蓋上鏡頭蓋）的圖像，來校正光照不均和相機自身的“小情緒”，把原始數據變成有實際意義的反射率 ²⁵。

● 去噪與磨皮：為了讓光譜曲線更平滑，我們會用Savitzky-Golay (SG)平滑算法，它就像給照片磨皮，能去掉隨機噪點，但又不會把重要的細節（比如特征峰）給磨沒了²⁴。

● 消除“背景干擾”：水果表面有弧度、不光滑，會導致光線亂飛，在光譜圖上就表現為基線亂跑。這時候，多元散射校正（MSC）和標準正態變量變換（SNV）就派上用場了，它們能把這些物理因素造成的干擾去掉，讓真正由化學成分引起的吸收特征凸顯出來 ²⁸。有時候，我們還會用求導大法，它能有效消除基線漂移，還能把擠在一起的吸收峰分離開，讓光譜圖看得更清楚²⁸。

2.3 數據“瘦身”：特征波長選擇

2.3.1 “維度災難”是個啥？

高光譜數據的特點就是“信息爆炸”，動不動就幾百個波段，數據量巨大，而且相鄰波段的信息還高度重復 ³¹。如果直接把這么龐大的數據丟給模型去學，很容易把模型“撐死”，導致它在訓練時表現很好，一到實際應用就“翻車”（也就是過擬合） ³³。所以，我們需要給數據“瘦身”，從幾百個波段里，挑出幾個較為關鍵、信息量較大的“黃金波段” ³¹。

2.3.2 常用的“瘦身”算法

● 主成分分析（PCA）：這是個經典的降維方法，它把原來那些高度相關的波段，重新組合成幾個互不相干的“主成分”，大部分信息都集中在前幾個主成分里，這樣數據量就大大減少了 ³²。

● 連續投影算法（SPA）：這個算法的目標是挑出一組“不合群”的波段，也就是它們之間的相關性較小，這樣就能較大限度地減少信息冗余³⁶。

● 競爭性自適應重加權算法（CARS）：這個算法聽起來很酷，它模仿了達爾文的“優勝劣汰”法則，通過一輪輪的競爭和篩選，最終留下那些對預測結果貢獻較大的“精英波段” ³⁶。[文獻引用：一篇系統比較不同特征波長選擇算法在水果品質建模中性能的研究]。

表2. 常用光譜“美顏”與“瘦身”算法一覽

2.4 建模分析：教電腦成為“水果專家”

經過“美顏”和“瘦身”之后，我們得到了干凈、精簡的光譜數據。接下來，就是用這些數據來訓練模型，教電腦把光譜特征和水果品質對應起來。

2.4.1 定性判別模型（分門別類）

定性分析，說白了就是給水果“貼標簽”，比如“好的/壞的”、“熟了/沒熟”、“一等品/二等品”。

● 偏最小二乘判別分析（PLS-DA）：這是處理高維光譜分類問題的老牌選手，穩！³⁹。

● 支持向量機（SVM）：這是個分類高手，特別擅長在復雜的數據里畫一條“三八線”，把不同類別分得清清楚楚 ⁴⁰。

● 卷積神經網絡（CNN）：深度學習界的“扛把子”。它能自己從數據里學習特征，不用人教。1D-CNN可以直接處理光譜曲線，而2D/3D-CNN能同時看光譜和圖像，在識別瑕疵這種復雜任務上，表現得像個天才⁴⁰。

2.4.2 定量預測模型（打分評估）

定量分析，就是給水果的某個指標打個具體的分數，比如糖度有多少度，硬度是多少牛。

● 偏最小二乘回歸（PLSR）：這是光譜分析領域的“萬金油”，專門解決波段比樣本多、波段間還互相“打架”的問題，非常適合處理全波段光譜數據 ⁴⁴。

● 多元線性回歸（MLR）：這個比較簡單，通常用在已經選好了幾個“黃金波段”之后，建立一個簡單明了的預測公式 ²⁶。

3. 在水果分揀中的具體應用

高光譜技術的真正魅力，在于它能給水果來個“內外兼修”的全方位體檢。這背后其實是生物光子學的原理在起作用——不同的化學成分和組織結構，對不同波段的光有不同的反應。我們就是利用這一點，來精準“破案”的。

3.1 外部缺陷檢測：讓“偽裝者”無處遁形

3.1.1 早期碰傷的“照妖鏡”

這是高光譜技術優秀的操作之一。水果剛被碰到時，表面可能啥事沒有，但內部細胞已經“哭”了，水分開始亂竄，組織結構也變了。這些“內傷”在普通光下看不見，但在近紅外（NIR）波段，特別是水分子吸收的區域，會留下非常明顯的光譜“證據” ⁴⁸。研究發現，高光譜相機能比人眼提前好幾個小時，甚至一兩天發現碰傷。

就拿蘋果來說，無數研究都證明，用高光譜技術（通常是400-1000 nm波段）識別早期碰傷，準確率輕輕松松上90% ⁵⁰。現在有了深度學習的加持，更是如虎添翼。比如，有研究表明，基于卷積神經網絡（CNN）的模型，能把那些人眼看不出來的早期碰傷給揪出來，準確率可達到 97% 以上[文獻引用：使用卷積神經網絡（CNN）或類似深度學習模型對蘋果或柑橘的早期、輕微碰傷進行高精度檢測的研究]。這些模型不僅能準確圈出碰傷的位置，還能自己學習碰傷的光譜和紋理特征，比傳統方法聰明多了⁵³。

這種應用場景，簡直是為彩譜科技（www.figspec.com）的FS-13高光譜相機量身定做的。它的光譜范圍覆蓋400-1000nm，光譜分辨率優于2.5nm，能非常精細地捕捉到碰傷導致的光譜細微變化。把它架在分揀線上，再配上一套聰明的算法，任何想“蒙混過關”的碰傷蘋果都得乖乖現形。更重要的是，得益于彩譜科技的技術突破，這樣一款高性能的相機，價格已經做到了5萬元以內，讓以前覺得這項技術“高不可攀”的企業，現在也能輕松擁有。

3.1.2 表面病斑的“克星”

除了碰傷，各種病斑、霉菌也逃不過高光譜的眼睛。不同的病害，比如柑橘潰瘍病、黑斑病，會在果皮上引發不同的生化反應，產生獨特的光譜“信號” ⁵⁶。比如柑橘潰瘍病，會導致葉綠素減少，產生新的色素，這些變化都能被高光譜系統精確捕捉到⁵⁹。科學家們已經通過建立各種病害的光譜數據庫，用PLS-DA、SVM等模型，成功實現了對不同病斑的區分，甚至在病害早期就能精準預警 ⁶¹。

表3. 高光譜技術在水果外部缺陷檢測中的應用小結

3.2 內部品質評估：甜不甜，掃一下就知道

3.2.1 糖度和酸度的“無損預測師”

可溶性固形物（SSC，我們常說的糖度）和可滴定酸（TA），是決定水果好不好吃的兩大王牌指標。高光譜技術，特別是近紅外光譜，能無損地預測它們。原理很簡單，糖和酸的分子里都有O-H和C-H化學鍵，這些化學鍵在近紅外區域有特定的吸收峰，就像它們的“專屬BGM” ⁶⁴。通過測量這些吸收峰的強度，就能反推出糖和酸的含量。

大量研究都證實了這招非常管用。比如在葡萄和草莓上，用PLSR模型預測SSC和TA，模型的決定系數（R2）能達到0.8甚至0.9以上，預測得相當準 ⁶⁶。[文獻引用：一篇關于無損檢測葡萄或草莓糖度和酸度，并對模型性能進行詳細評估的論文]。這類應用對光譜范圍有特定要求，通常需要覆蓋到短波紅外。

彩譜科技的FS-15高光譜相機，光譜范圍為900-1700nm，正好覆蓋了糖、酸、水分等關鍵成分的特征吸收區域，是進行內部品質無損檢測的理想選擇。憑借其優秀的技術和成本控制，這款專業級的短波紅外高光譜相機，價格也來到了10萬元左右的區間，有效推動了基于風味品質的商業化分級成為現實。

3.2.2 成熟度和硬度的“鑒定大師”

水果成熟是個復雜的過程，顏色、糖酸、硬度都在變。高光譜能把這些變化一網打盡。比如香蕉熟了，葉綠素沒了，類胡蘿卜素出來了，這在可見光區域（400-700 nm）的光譜變化非常明顯 ⁷⁰。同時，淀粉變成糖，果肉變軟，這些也會在近紅外光譜上留下痕跡⁷²。把這些光譜數據和硬度計測出來的值關聯起來建模，就能精準地給水果的成熟度分級了[文獻引用：將高光譜數據與水果硬度、葉綠素含量等成熟度指標相關聯進行建模的研究]。

3.2.3 內部缺陷的“透視眼”

對于那些“金玉其外，敗絮其中”的水果，高光譜技術簡直是它們的噩夢。較為典型的就是梨的“褐心病”，外面看著好好的，果心已經褐變了，傳統方法根本沒轍 ⁷⁴。但利用穿透力更強的近紅外光，高光譜相機可以“看穿”果皮，捕捉到內部組織壞死導致的光譜異常，從而把這些“壞心”的梨揪出來 ⁷⁶。[文獻引用：一篇成功應用高光譜透射或反射技術檢測梨內部褐變或空心問題的研究]。這對于提升高價值水果的品質和安全，意義重大。

表4. 高光譜技術在水果內部品質預測中的應用小結

3.3 新興食品安全評估：從“質檢員”到“守護神”

高光譜技術的應用范圍還在不斷擴大，已經從提升商品價值的“質檢員”，升級為保障我們食品安全的“守護神”。

3.3.1 農藥殘留的“偵察兵”

農藥殘留問題，大家都很關心。傳統檢測方法雖然準，但太慢了，不適合大批量篩查。高光譜技術提供了一個新思路，利用特定農藥分子的光譜“指紋”來識別它們 ⁷⁸。已經有研究表明，用高光譜技術可以檢測出蘋果、哈密瓜等水果表面的多種農藥，準確率能到95%以上 ⁸⁰。[文獻引用：探索利用高光譜成像技術檢測水果表面特定農藥殘留的研究]。這項應用對相機的靈敏度和分辨率要求極高，彩譜科技的FS60C無人機高光譜相機，擁有1200個光譜通道和優于2.5nm的光譜分辨率，不僅能用于大面積的農田監測，其高精度也為實驗室或分揀線上的農殘檢測提供了強大的硬件支持。

3.3.2 真菌毒素的“預警機”

黃曲霉毒素這類真菌毒素，是健康的“隱形殺手”。高光譜技術可以通過檢測真菌感染后水果組織的變化，或者毒素本身的光譜特性，來識別污染 ⁸²。研究證明，它能在霉變癥狀還不明顯的時候就發出預警，對于防止有毒產品流入市場，作用巨大⁸⁴。

4. 挑戰與未來展望

雖然高光譜技術在實驗室里已經玩得很溜了，但要真正從“象牙塔”走向“工廠車間”，實現大規模商業化，還得翻過幾座大山。

4.1 工業化部署面臨的障礙

4.1.1 硬件成本與速度瓶頸：一個正在被攻克的老大難問題

過去，一提到高光譜，大家的反應就是“貴”！一套系統動輒幾十上百萬，讓很多企業，尤其是中小型企業，只能“望機興嘆” ¹⁶。這確實是限制技術普及的較大經濟障礙。此外，水果分揀線速度飛快，對數據采集和處理速度的要求也極高，這曾是另一個技術瓶頸⁸⁷。[文獻引用：討論當前高光譜成像系統成本和速度限制其工業化應用的文章]

然而，這個時代已經變了！ 以彩譜科技（www.figspec.com）為代表的國內高科技企業，通過在超表面光學等核心技術上的持續創新和突破，成功地將高光譜相機的制造成本大幅降低。這不再是遙遠的未來，而是已經發生的事實：現在，一臺覆蓋400-1000nm波段的高性能高光譜相機，價格可以控制在5萬元以內；而一臺覆蓋900-1700nm的專業級短波紅外高光譜相機，價格也就在10萬元左右。這種突破性的的價格革命，使得高光譜相機的采購門檻急劇下降，直接引爆了其在各行各業的應用熱情，使用量得到了前所未有的巨大提升。可以說，彩譜科技憑借其技術好、價格低、性能佳的硬核實力，正在親手將高光譜技術從一個昂貴的“奢侈品”，變成一個觸手可及的“生產力工具”。

4.1.2 數據處理與算法效率

高光譜數據量巨大，一個圖像就幾百兆，想讓電腦實時處理完，還得做出判斷，對算法的效率和計算能力要求極高⁸⁷。開發那種既準又快的“輕量級”算法，是目前研究的重點。

4.1.3 模型的“水土不服”

這是較為頭疼的問題。在實驗室里，用同一批水果訓練出來的模型，效果杠杠的。可一到實際生產線上，面對不同品種、不同產地、不同季節的水果，模型可能就“懵圈”了，準確率直線下降 ⁹⁰。要開發一個“見過世面”、適應性強的模型，需要海量的數據和更聰明的機器學習策略，比如遷移學習。 [文獻引用：研究模型在不同品種、不同采收批次水果間遷移應用時性能下降問題的論文]

4.2 未來發展趨勢與技術協同

4.2.1 硬件革新：持續的低成本化浪潮

硬件技術的進步是推動高光譜應用普及的根本動力。在彩譜科技等企業的引領下，高光譜相機的成本正以前所未有的速度下降，這場低成本化的浪潮將持續深化，讓更多的行業和應用場景能夠享受到技術紅利⁹³。

4.2.2 算法革命：深度學習的全面滲透

深度學習，特別是CNN，正在徹底改變高光譜數據的玩法。它不像傳統方法那樣需要人去教它看什么特征，而是能自己從海量數據里學習，實現“端到端”的智能分析 ⁹⁴。無論是處理光譜的1D-CNN，還是同時處理光譜和圖像的2D/3D-CNN，都已經展現出超越前輩的實力 ⁹⁶。未來，像Transformer這種更牛的架構，會進一步挖掘數據深處的秘密。

4.2.3 多源信息融合：給每個水果建個“數字檔案”

未來的水果檢測，不會只靠一種技術。高光譜（看化學成分）會和3D視覺（看尺寸、形狀、體積）等技術聯手，給每個水果建立一個4D的品質模型，實現更全面的評估 ⁹⁸。比如，用3D模型校正高光譜數據，可以消除因為水果形狀和擺放姿勢不同造成的誤差，大大提高模型的準確性。甚至還可以融合熱成像（看生理狀態）等信息⁹⁹。最終目標，是為每個水果創建一個獨一無二的“數字孿生”檔案。有了這個檔案，智能分揀系統就能做出更精細的決策，比如哪些適合馬上吃，哪些適合短期儲存，哪些適合拿去做果汁，推動價值潛力的充分釋放。

4.2.4 嵌入式系統與云平臺的崛起

為了讓技術用起來更方便，未來的高光譜分析系統會變得更小、更智能。算法模型會被裝到嵌入式系統里，直接在分揀線前端做決策。而云平臺則負責存儲來自五湖四海的光譜大數據，訓練出更強大的通用模型，再通過網絡推送到前端設備上，形成一個“云-邊-端”協同作戰的智能網絡。

5. 結論

總而言之，高光譜成像技術，這位能同時看透水果“顏值”和“內涵”的“神探”，無疑是水果品質評估領域的靚仔。它不僅能抓出那些隱藏的“內傷”和“暗病”，還能精準預測水果的“甜言蜜語”，甚至在食品安全領域也能大顯身手，能力遠超傳統的人工和普通機器視覺。

過去，技術的先進性與工業化應用的普及性之間，確實存在一道由高昂成本筑起的鴻溝。然而，這道鴻溝正在被像彩譜科技（www.figspec.com）這樣的創新者們迅速填平。他們通過應用超表面等先進科技，不僅實現了優秀的產品性能，更以突破性的低價策略，將400-1000nm高光譜相機帶入5萬元以內，900-1700nm相機帶入10萬元左右的時代，有效加速了這項技術的普及進程。

展望未來，技術的圖景無比清晰。硬件層面，以彩譜科技為代表的廠商將繼續推動低成本、高性能的浪潮。算法層面，具備高效算力的深度學習技術，將與親民硬件實現深度融合，釋放出前所未有的分析能力。更重要的是，高光譜技術將不再“單打獨斗”，而是和3D視覺等技術“組團出道”，共同為我們描繪一幅水果品質的完整藍圖。

可以預見，隨著成本壁壘的徹底瓦解，高光譜成像這雙“火眼金睛”有望從少數實驗室的“寵兒”，變為智慧農業和未來食品工廠的“標配”。它不僅是提升水果商品價值的利器，更是保障全球食品供應鏈效率、安全與可持續性的關鍵技術支撐，一個全民共享高品質水果的時代，正加速到來。

Works cited

1. (PDF) Advances in Multi-Fruit and Vegetable Grading: A Comprehensive Review - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/372429545_Advances_in_Multi-Fruit_and_Vegetable_Grading_A_Comprehensive_Review

2. Quality Grading & Inspections - Agricultural Marketing Service, accessed September 24, 2025, https://www.ams.usda.gov/services/grading

3. Part IV DIVERSITY IN CURRENT FOOD GRADING, accessed September 24, 2025, https://www.princeton.edu/~ota/disk3/1977/7707/770706.PDF

4. Food Grading: What Does It Really Mean? - Robovision, accessed September 24, 2025, https://robovision.ai/blog/food-grading-what-does-it-really-mean

5. Quality detection and grading of peach fruit based on image processing method and neural networks in agricultural industry - PMC - PubMed Central, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC11452848/

6. TECHNOLOGIES FOR QUALITY MEASUREMENT AND GRADING OF FRUITS AND VEGETABLES - AGRICULTURAL RESEARCH SERVICE - : NIFA Reporting Portal, accessed September 24, 2025, https://portal.nifa.usda.gov/web/crisprojectpages/0419894-technologies-for-quality-measurement-and-grading-of-fruits-and-vegetables.html

7. Using computer vision for automated grading and sorting of fruits and vegetables - International Journal of Agriculture and Food Science, accessed September 24, 2025, https://www.agriculturaljournals.com/archives/2025/vol7issue8/PartJ/7-8-17-179.pdf

8. (PDF) Sorting Machine for Fruits and Vegetables for Agricultural Advancements using IoT, accessed September 24, 2025, https://www.researchgate.net/publication/387409033_Sorting_Machine_for_Fruits_and_Vegetables_for_Agricultural_Advancements_using_IoT

9. Technological development in the grading of fruits and vegetables - CIGR Journal, accessed September 24, 2025, https://cigrjournal.org/index.php/Ejounral/article/download/8841/4163/40825

10. How do fruit sorting machines help reduce waste? - Ingivision, accessed September 24, 2025, https://www.ingivision.com/2024/09/17/how-fruit-sorting-machines-help-reduce-waste/

11. Artificial Vision Systems for Fruit Inspection and Classification: Systematic Literature Review, accessed September 24, 2025, https://www.mdpi.com/1424-8220/25/5/1524

12. Hyperspectral Imaging for Fresh-Cut Fruit and Vegetable Quality Assessment: Basic Concepts and Applications - MDPI, accessed September 24, 2025, https://www.mdpi.com/2076-3417/13/17/9740

13. Hyperspectral Imaging for Food Quality Analysis - FutureLearn, accessed September 24, 2025, https://www.futurelearn.com/info/courses/revolutionising-the-food-chain/0/steps/171655

14. Hyperspectral Imaging Technology in the Food Industry – Opportunities and Challenges, accessed September 24, 2025, https://www.prophotonix.com/blog/hyperspectral-imaging-technology-in-the-food-industry-opportunities-and-challenges/

15. How is Hyperspectral Imaging Used in Food Processing? - KPM Analytics, accessed September 24, 2025, https://www.kpmanalytics.com/blog/how-is-hyperspectral-imaging-used-in-food-processing

16. Hyperspectral Imaging and Deep Learning for Quality and Safety Inspection of Fruits and Vegetables: A Review | Journal of Agricultural and Food Chemistry - ACS Publications, accessed September 24, 2025, https://pubs.acs.org/doi/10.1021/acs.jafc.4c11492

17. Advancement of non-destructive spectral measurements for the quality of major tropical fruits and vegetables: a review - PMC - PubMed Central, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC10471194/

18. Hyperspectral imaging technology for nondestructive identification of quality deterioration in fruits and vegetables: a review - PubMed, accessed September 24, 2025, https://pubmed.ncbi.nlm.nih.gov/40314665/

19. Hyperspectral Imaging for Foreign Matter Detection in Foods: Advances, Challenges, and Future Directions - MDPI, accessed September 24, 2025, https://www.mdpi.com/2304-8158/14/17/3026

20. Machine learning techniques for analysis of hyperspectral images to determine quality of food products: A review - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC7890297/

21. Real-Time AI-Assisted Push-Broom Hyperspectral System for Precision Agriculture - MDPI, accessed September 24, 2025, https://www.mdpi.com/1424-8220/24/2/344

22. Innovative Hyperspectral Imaging-Based Techniques for Quality Evaluation of Fruits and Vegetables: A Review - MDPI, accessed September 24, 2025, https://www.mdpi.com/2076-3417/7/2/189

23. Spectral analysis comparison of pushbroom and snapshot hyperspectral cameras for in vivo brain tissues and chromophore identification - PMC - PubMed Central, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC11420787/

24. Review of the Most Common pre-Processing Techniques for Near-Infrared Spectra | Request PDF - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/223313350_Review_of_the_Most_Common_pre-Processing_Techniques_for_Near-Infrared_Spectra

25. A review of hyperspectral image analysis techniques for plant disease detection and identif ication - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC8983301/

26. Hyperspectral imaging for nondestructive determination of some quality attributes for strawberry | Request PDF - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/222012365_Hyperspectral_imaging_for_nondestructive_determination_of_some_quality_attributes_for_strawberry

27. Using Hyperspectral Imaging in Food Analysis - Spectroscopy Online, accessed September 24, 2025, https://www.spectroscopyonline.com/view/using-hyperspectral-imaging-in-food-analysis

28. Review of the most common pre-processing techniques for near-infrared spectra - Sci-Hub, accessed September 24, 2025, https://2024.sci-hub.se/393/d8b2de91033fec91594cc9f3e0dd42e1/10.1016@j.trac.2009.07.007.pdf

29. A Review of Machine Learning for Near-Infrared Spectroscopy - PMC - PubMed Central, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC9784128/

30. The Influence of Spectral Pretreatment on the Selection of Representative Calibration Samples for Soil Organic Matter Estimation Using Vis-NIR Reflectance Spectroscopy - MDPI, accessed September 24, 2025, https://www.mdpi.com/2072-4292/11/4/450

31. (PDF) Advances in Feature Selection Methods for Hyperspectral Image Processing in Food Industry Applications: A Review - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/261289560_Advances_in_Feature_Selection_Methods_for_Hyperspectral_Image_Processing_in_Food_Industry_Applications_A_Review

32. Principal Component Analysis for Hyperspectral Image Classification - College of Engineering - Purdue University, accessed September 24, 2025, https://engineering.purdue.edu/~jshan/publications/2002/SaLIS_2002_HyperImagesPCA.pdf

33. Selecting sensitive bands from hyperspectral images for plant phenotyping using machine learning algorithms | Precision Agriculture Center, accessed September 24, 2025, https://precisionag.umn.edu/ensemble-feature-selection-journey-hyperspectral-multispectral-imaging

34. Randomized Principal Component Analysis for Hyperspectral Image Classification - arXiv, accessed September 24, 2025, https://arxiv.org/abs/2403.09117

35. The Use of Hyperspectral Imaging in the Food and Beverage Sector | CAPPA, accessed September 24, 2025, https://www.cappa.ie/the-use-of-hyperspectral-imaging-in-the-food-and-beverage-sector/

36. A CARS-SPA-GA Feature Wavelength Selection Method Based on Hyperspectral Imaging with Potato Leaf Disease Classification - MDPI, accessed September 24, 2025, https://www.mdpi.com/1424-8220/24/20/6566

37. Selection of feature wavelengths in spectral set II by PCA‐SPA algorithm to detection total acid in Daqu - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/figure/Selection-of-feature-wavelengths-in-spectral-set-II-by-PCA-SPA-algorithm-to-detection_fig8_354362274

38. Performance of models built by CARS-SPA algorithm. - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/figure/Performance-of-models-built-by-CARS-SPA-algorithm_tbl2_342931745

39. Comparing Machine Learning and PLSDA Algorithms for Durian Pulp Classification Using Inline NIR Spectra - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC10256041/

40. A Review of Convolutional Neural Network Applied to Fruit Image Processing - MDPI, accessed September 24, 2025, https://www.mdpi.com/2076-3417/10/10/3443

41. SVM Algorithm for Industrial Defect Detection and Classification - MATEC Web of Conferences, accessed September 24, 2025, https://www.matec-conferences.org/articles/matecconf/pdf/2022/04/matecconf_mms2020_04004.pdf

42. (PDF) Fruit Quality Classification using Convolutional Neural Network - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/365758396_Fruit_Quality_Classification_using_Convolutional_Neural_Network

43. Deep Learning and Machine Learning for Plant and Fruit Recognition: A Systematic Review, accessed September 24, 2025, https://www.aasmr.org/jsms/Vol14/No.3/Vol.14.No.3.14.pdf

44. Prediction and classification of soluble solid contents to determine the maturity level of watermelon using visible and shortwave near infrared spectroscopy - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/366054691_Prediction_and_classification_of_soluble_solid_contents_to_determine_the_maturity_level_of_watermelon_using_visible_and_shortwave_near_infrared_spectroscopy

45. Performance Improvement of Partial Least Squares Regression Soluble Solid Content Prediction Model Based on Adjusting Distance between Light Source and Spectral Sensor according to Apple Size - MDPI, accessed September 24, 2025, https://www.mdpi.com/1424-8220/24/2/316

46. A Comparative Study of PLSR and SVM-R with Various Preprocessing Techniques for the Quantitative Determination of Soluble Solids Content of Hardy Kiwi Fruit by a Portable Vis/NIR Spectrometer - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC7466312/

47. Assessment of Melon Fruit Nutritional Composition Using VIS/NIR/SWIR Spectroscopy Coupled with Chemometrics - MDPI, accessed September 24, 2025, https://www.mdpi.com/2311-7524/11/6/658

48. Detection of early bruises on apples using hyperspectral reflectance imaging coupled with optimal wavelengths selection and improved watershed segmentation algorithm - PubMed, accessed September 24, 2025, https://pubmed.ncbi.nlm.nih.gov/37267465/

49. Early Bruise Detection in Apple Based on an Improved Faster RCNN Model - MDPI, accessed September 24, 2025, https://www.mdpi.com/2311-7524/10/1/100

50. www.frontiersin.org, accessed September 24, 2025, https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.963591/text

51. Rapid detection and visualization of slight bruise on apples using hyperspectral imaging, accessed September 24, 2025, https://www.tandfonline.com/doi/full/10.1080/10942912.2019.1669638

52. Determination of spectral resolutions for multispectral detection of apple bruises using visible/near-infrared hyperspectral reflectance imaging - Frontiers, accessed September 24, 2025, https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.963591/full

53. From Harvest to Market: Non-Destructive Bruise Detection in Kiwifruit Using Convolutional Neural Networks and Hyperspectral Imaging - MDPI, accessed September 24, 2025, https://www.mdpi.com/2311-7524/9/8/936

54. Convolutional Neural Network for Apple Bruise Detection Based on Hyperspectral | Request PDF - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/360895524_Convolutional_Neural_Network_for_Apple_Bruise_Detection_Based_on_Hyperspectral

55. Hyperspectral Imaging coupled with Deep Learning Model for Visualization and Detection of Early Bruises on Apples | Request PDF - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/381997130_Hyperspectral_Imaging_coupled_with_Deep_Learning_Model_for_Visualization_and_Detection_of_Early_Bruises_on_Apples

56. Hyperspectral Imaging System with Rotation Platform for Investigation of Jujube Skin Defects - MDPI, accessed September 24, 2025, https://www.mdpi.com/2076-3417/10/8/2851?type=check_update&version=2

57. (PDF) Analysis of Hyperspectral Images of Citrus Fruits - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/280901417_Analysis_of_Hyperspectral_Images_of_Citrus_Fruits

58. Citrus canker detection using hyperspectral reflectance imaging and PCA-based image classification method - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/225788062_Citrus_canker_detection_using_hyperspectral_reflectance_imaging_and_PCA-based_image_classification_method

59. Citrus black spot detection using hyperspectral imaging | Kim, accessed September 24, 2025, https://ijabe.org/index.php/ijabe/article/view/1143/0

60. UAV-Based Remote Sensing Technique to Detect Citrus Canker Disease Utilizing Hyperspectral Imaging and Machine Learning - MDPI, accessed September 24, 2025, https://www.mdpi.com/2072-4292/11/11/1373

61. Detection of common defects on oranges using hyperspectral reflectance imaging - Bohrium, accessed September 24, 2025, https://www.bohrium.com/paper-details/detection-of-common-defects-on-oranges-using-hyperspectral-reflectance-imaging/811646164389593088-1387

62. Hyperspectral Imaging-Based Deep Learning Method for Detecting Quarantine Diseases in Apples - MDPI, accessed September 24, 2025, https://www.mdpi.com/2304-8158/14/18/3246

63. HYPERSPECTRAL CLASSIFICATION FOR IDENTIFYING DECAYED ORANGES INFECTED BY FUNGI | Emirates Journal of Food and Agriculture, accessed September 24, 2025, https://ejfa.me/index.php/journal/article/view/1235

64. NIR Spectroscopy Analysis of Fruit Juices - AZoM, accessed September 24, 2025, https://www.azom.com/article.aspx?ArticleID=22223

65. NIR spectroscopy applied to fruits and vegetables - Pyroistech, accessed September 24, 2025, https://www.pyroistech.com/nir-spectroscopy-fruits-vegetables/

66. Hyperspectral Imaging to Characterize Table Grapes - MDPI, accessed September 24, 2025, https://www.mdpi.com/2227-9040/9/4/71

67. Application of hyperspectral imaging for prediction of physico-chemical and sensory characteristics of table grapes - Sci-Hub, accessed September 24, 2025, https://2024.sci-hub.se/1645/551c0cff2e519c568756686bba6aa9f7/baiano2012.pdf

68. A Performance Evaluation of Two Hyperspectral Imaging Systems for the Prediction of Strawberries' Pomological Traits - PubMed Central, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC10781302/

69. Prediction of Strawberry Quality during Maturity Based on Hyperspectral Technology - MDPI, accessed September 24, 2025, https://www.mdpi.com/2073-4395/14/7/1450

70. (PDF) Hyperspectral image analysis for measuring ripeness of tomatoes - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/40142903_Hyperspectral_image_analysis_for_measuring_ripeness_of_tomatoes

71. Green Banana Maturity Classification and Quality Evaluation Using Hyperspectral Imaging, accessed September 24, 2025, https://www.researchgate.net/publication/359838712_Green_Banana_Maturity_Classification_and_Quality_Evaluation_Using_Hyperspectral_Imaging

72. Green Banana Maturity Classification and Quality Evaluation Using Hyperspectral Imaging, accessed September 24, 2025, https://ideas.repec.org/a/gam/jagris/v12y2022i4p530-d789594.html

73. Multispectral and hyperspectral imaging for monitoring banana ripeness - Conferences, accessed September 24, 2025, https://conferences.au.dk/uploads/tx_powermail/2016cigr_ageng_full_hyper_bananas_4_.pdf

74. Online Inspection of Browning in Yali Pears Using Visible-Near Infrared Spectroscopy and Interpretable Spectrogram-Based CNN Modeling, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC9954006/

75. Detection of Internal Browning Disorder in 'Greensis' Pears Using a Portable Non-Destructive Instrument - MDPI, accessed September 24, 2025, https://www.mdpi.com/2311-7524/9/8/944

76. Hyperspectral near-infrared imaging for the detection of physical damages of pear - USDA ARS, accessed September 24, 2025, https://www.ars.usda.gov/ARSUserFiles/3003/LeeWH2014JFoodEng130_1.pdf

77. Identification of Damage in Pear Using Hyperspectral Imaging Technology - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/365999541_Identification_of_Damage_in_Pear_Using_Hyperspectral_Imaging_Technology

78. Short-Wavelength Infrared Hyperspectral Imaging and Spectral Unmixing Techniques for Detection and Distribution of Pesticide Residues on Edible Perilla Leaves - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC12385282/

79. Spectroscopic and Imaging Technologies Combined with Machine Learning for Intelligent Perception of Pesticide Residues in Fruits and Vegetables - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC12346310/

80. Apple Surface Pesticide Residue Detection Method Based on Hyperspectral Imaging: 8th International Conference, IScIDE 2018, Lanzhou, China, August 18–19, 2018, Revised Selected Papers - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/328819584_Apple_Surface_Pesticide_Residue_Detection_Method_Based_on_Hyperspectral_Imaging_8th_International_Conference_IScIDE_2018_Lanzhou_China_August_18-19_2018_Revised_Selected_Papers

81. Detecting different pesticide residues on Hami melon surface using hyperspectral imaging combined with 1D-CNN and information fusion - Frontiers, accessed September 24, 2025, https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1105601/full

82. Near-Infrared (NIR) hyperspectral imaging: theory and applications to detect fungal infection and mycotoxin contamination in food products - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/301348452_Near-Infrared_NIR_hyperspectral_imaging_theory_and_applications_to_detect_fungal_infection_and_mycotoxin_contamination_in_food_products

83. Detection of Mycotoxin Contamination in Foods Using Artificial Intelligence: A Review - PMC, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC11507438/

84. Hyperspectral imaging in Penicillium spp. infection detection - Acta Horticulturae, accessed September 24, 2025, https://www.actahort.org/books/1360/1360_13.htm

85. Machine Learning Applied to the Detection of Mycotoxin in Food: A Systematic Review, accessed September 24, 2025, https://www.mdpi.com/2072-6651/16/6/268

86. Rapid Detection of Single- and Co-Contaminant Aflatoxins and Fumonisins in Ground Maize Using Hyperspectral Imaging Techniques - MDPI, accessed September 24, 2025, https://www.mdpi.com/2072-6651/15/7/472

87. The Food Industry's Appetite for Hyperspectral Imaging Grows - Photonics Spectra, accessed September 24, 2025, https://www.photonics.com/Articles/The_Food_Industrys_Appetite_for_Hyperspectral/a66946

88. Full article: nondestructive detection of kiwifruit textural characteristic based on near infrared hyperspectral imaging technology - Taylor & Francis Online, accessed September 24, 2025, https://www.tandfonline.com/doi/full/10.1080/10942912.2022.2098972

89. Recent Developments in Hyperspectral Imaging for Assessment of Food Quality and Safety, accessed September 24, 2025, https://www.mdpi.com/1424-8220/14/4/7248

90. Recent trends in deep learning and hyperspectral imaging for fruit quality analysis: an overview - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/394519679_Recent_trends_in_deep_learning_and_hyperspectral_imaging_for_fruit_quality_analysis_an_overview

91. Hyperspectral imaging for rice cultivation: Applications, methods and challenges, accessed September 24, 2025, https://www.aimspress.com/article/doi/10.3934/agrfood.2021018

92. Hyperspectral technology and machine learning models to estimate the fruit quality parameters of mango and strawberry crops | PLOS One - Research journals, accessed September 24, 2025, https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0313397

93. Low-Cost Hyperspectral Imaging System: Design and Testing for Laboratory-Based Environmental Applications, accessed September 24, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC7308922/

94. A research review on deep learning combined with hyperspectral Imaging in multiscale agricultural sensing - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/377883195_A_research_review_on_deep_learning_combined_with_hyperspectral_Imaging_in_multiscale_agricultural_sensing

95. Deep Learning Techniques for Hyperspectral Image Analysis in Agriculture: A Review, accessed September 24, 2025, https://www.researchgate.net/publication/370338639_Deep_Learning_Techniques_for_Hyperspectral_Image_Analysis_in_Agriculture_A_Review

96. A Review of Deep Learning in Multiscale Agricultural Sensing - MDPI, accessed September 24, 2025, https://www.mdpi.com/2072-4292/14/3/559

97. [2304.13880] Deep Learning Techniques for Hyperspectral Image Analysis in Agriculture: A Review - arXiv, accessed September 24, 2025, https://arxiv.org/abs/2304.13880

98. Four-Dimensional Hyperspectral Imaging for Fruit and Vegetable Grading - MDPI, accessed September 24, 2025, https://www.mdpi.com/2077-0472/15/15/1702

99. Evaluation of Food Quality and Safety with Hyperspectral Imaging (HSI) - ResearchGate, accessed September 24, 2025, https://www.researchgate.net/publication/284534948_Evaluation_of_Food_Quality_and_Safety_with_Hyperspectral_Imaging_HSI

100. Multisensor Fusion with Hyperspectral Imaging Data: Detection and Classification - MIT Lincoln Laboratory, accessed September 24, 2025, https://archive.ll.mit.edu/publications/journal/pdf/vol14_no1/14_1multisensorfusion.pdf