TWI805513B - Intelligent bottle recycling identification method - Google Patents
Intelligent bottle recycling identification method Download PDFInfo
- Publication number
- TWI805513B TWI805513B TW111139935A TW111139935A TWI805513B TW I805513 B TWI805513 B TW I805513B TW 111139935 A TW111139935 A TW 111139935A TW 111139935 A TW111139935 A TW 111139935A TW I805513 B TWI805513 B TW I805513B
- Authority
- TW
- Taiwan
- Prior art keywords
- recycled
- image
- item
- processing module
- reference area
- Prior art date
Links
Images
Landscapes
- Processing And Handling Of Plastics And Other Materials For Molding In General (AREA)
- Processing Of Solid Wastes (AREA)
Abstract
本發明係提供一種智慧瓶型回收辨識方法,係以一回收機台執行,該回收機台設置有一影像擷取模組及一處理模組,該智慧瓶型回收辨識方法包括以下步驟:影像擷取模組擷取待回收物品之影像並取得有一原始影像;處理模組接收所述原始影像且對原始影像之背景進行裁切並取得有一首處理影像;再對首處理影像中之待回收物品進行邊緣檢測並取得一邊緣影像且於該邊緣影像外圍界定有一定位框,與將所述定位框形成於所述首處理影像上;於該首處理影像之定位框內界定有一第一基準區域及一第二基準區域且對第一基準區域與第二基準區域內之影像判斷該待回收物品方向是否正確;待回收物品方向正確時,該處理單元再對首處理影像中之定位框內界定有至少一局部框,由局部框內之顏色判定該待回收物品之顏色且以一顏色設定條件判斷該待回收物品是否進行回收;及待回收物品顏色符合顏色條件設定時,再以一人工智慧模型對該待回收物品進行辨識,藉此達到可精準判斷回收物以節省人力與時間耗費之功效者。The present invention provides a method for recycling and identifying smart bottle shapes, which is executed by a recycling machine. The recycling machine is provided with an image capture module and a processing module. The method for recycling and identifying smart bottle shapes includes the following steps: image capture The fetching module captures the image of the article to be recycled and obtains an original image; the processing module receives the original image and cuts the background of the original image to obtain a processed image; and then processes the article to be recycled in the first processed image Perform edge detection and obtain an edge image and define a positioning frame around the edge image, and form the positioning frame on the first processed image; define a first reference area in the positioning frame of the first processed image and A second reference area and judging whether the direction of the object to be recycled is correct based on the images in the first reference area and the second reference area; when the direction of the object to be recycled is correct, the processing unit defines the positioning frame in the first processed image. At least one partial frame, the color of the item to be recycled is judged by the color in the partial frame, and a color setting condition is used to determine whether the item to be recycled is recycled; and when the color of the item to be recycled meets the color condition setting, an artificial intelligence model is used Identify the items to be recycled, so as to achieve the effect of accurately judging the recycled items and saving manpower and time.
Description
本發明係有關於一種回收方法,尤指一種可精準判斷回收物以節省人力與時間耗費之智慧瓶型回收辨識方法。The invention relates to a recycling method, in particular to a smart bottle type recycling identification method that can accurately judge recycled objects to save manpower and time.
隨著生活品質不斷提升,民眾越加重視環境的保護,而環境的保護最主要就是廢棄物的分類回收,將不同的廢棄物分類收集再利用,已分解再製為環保產品,以達到資源再利用且減少原物料的消耗。With the continuous improvement of the quality of life, people pay more and more attention to environmental protection, and the most important thing in environmental protection is the classification and recycling of waste. Different wastes are collected and reused by classification, and have been decomposed and remanufactured into environmentally friendly products to achieve resource reuse. And reduce the consumption of raw materials.
政府有公告台灣在2025年要淨零碳排,針對目前 將一般回收垃圾丟垃圾桶,而後便需要請專人來整理並回收,所以在這些回收過程相對會產生運輸的人力成本,並且在回收的過程也會產生許多碳足跡,因為在回收處理廠一般回收垃圾時需要透過人力或是機器的分選,而在機器分選的過程中會產生廢水、廢熱及會產生廢碳,而廢碳的減少對地球環境是很有幫助的。 The government has announced that Taiwan will have net-zero carbon emissions by 2025. Throw the general recycling garbage into the trash can, and then you need to hire a special person to sort it out and recycle it. Therefore, the labor cost of transportation will be relatively incurred in these recycling processes, and a lot of carbon footprint will also be generated in the recycling process, because the recycling plant usually recycles Garbage needs to be sorted by manpower or machines. During the process of machine sorting, waste water, waste heat and waste carbon will be generated. The reduction of waste carbon is very helpful to the global environment.
而塑膠瓶因為使用上便利,所以大多飲品都是以寶特瓶或塑膠杯來容置,但隨著使用量越大,其廢棄的量越多,由於塑膠具有不易分解的特性,使得塑膠的回收以及回收後的處理方式變得格外重要,而目前一般常見的回收方式通常是藉由人工進行分類回收,如此不僅造成人力上的耗損外,也非常耗費時間。Plastic bottles are convenient to use, so most beverages are stored in plastic bottles or plastic cups, but as the usage increases, the amount of waste will increase. Due to the characteristics of plastics that are not easy to decompose, the plastic Recycling and post-recycling processing methods have become extremely important, and the current common recycling method is usually to sort and recycle manually, which not only causes manpower loss, but also takes a lot of time.
是以,要如何解決上述習用之問題與缺失,即為本發明之發明人與從事此行業之相關廠商所亟欲研究改善之方向所在者。Therefore, how to solve the above-mentioned conventional problems and deficiencies is the direction that the inventor of the present invention and related manufacturers engaged in this industry want to research and improve urgently.
爰此,為有效解決上述之問題,本發明之主要目的在於提供一種可精準判斷回收物以節省人力與時間耗費之智慧瓶型回收辨識方法。Therefore, in order to effectively solve the above-mentioned problems, the main purpose of the present invention is to provide a smart bottle type recycling identification method that can accurately judge recycled objects and save manpower and time.
為達上述目的,本發明係提供一種智慧瓶型回收辨識方法,係以一回收機台執行,該回收機台設置有一影像擷取模組及一處理模組,該智慧瓶型回收辨識方法包括以下步驟:影像擷取模組擷取待回收物品之影像並取得有一原始影像;處理模組接收所述原始影像且對原始影像之背景進行裁切並取得有一首處理影像;再對首處理影像中之待回收物品進行邊緣檢測並取得一邊緣影像且於該邊緣影像外圍界定有一定位框,與將所述定位框形成於所述首處理影像上;於該首處理影像之定位框內界定有一第一基準區域及一第二基準區域且對第一基準區域與第二基準區域內之影像判斷該待回收物品方向是否正確;待回收物品方向正確時,該處理單元再對首處理影像中之定位框內界定有至少一局部框,由局部框內之顏色判定該待回收物品之顏色且以一顏色設定條件判斷該待回收物品是否進行回收;及待回收物品顏色符合顏色條件設定時,再以一人工智慧模型對該待回收物品進行辨識。In order to achieve the above purpose, the present invention provides a smart bottle type recycling identification method, which is implemented by a recycling machine. The recycling machine is equipped with an image capture module and a processing module. The smart bottle type recycling identification method includes The following steps: the image capture module captures the image of the article to be recycled and obtains an original image; the processing module receives the original image and cuts the background of the original image to obtain a processed image; and then processes the first processed image Perform edge detection on the items to be recycled and obtain an edge image and define a positioning frame on the periphery of the edge image, and form the positioning frame on the first processed image; define a positioning frame in the first processed image A first reference area and a second reference area, and judge whether the direction of the article to be recycled is correct based on the images in the first reference area and the second reference area; There is at least one partial frame defined in the positioning frame, the color of the article to be recycled is determined by the color in the partial frame, and a color setting condition is used to determine whether the article to be recycled is recycled; and when the color of the article to be recycled meets the color condition setting, then An artificial intelligence model is used to identify the item to be recycled.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述處理模組對首處理影像進行灰階處理與影像模糊處理,且將灰階處理及影像模糊處理後的影像進行邊緣檢測取得所述邊緣影像。The present invention also discloses a smart bottle type recycling identification method, wherein the processing module performs grayscale processing and image blurring processing on the first processed image, and performs edge detection on the image after grayscale processing and image blurring processing to obtain the edge image.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述邊緣檢測時同時進行雜訊濾除處理及像素膨脹處理。The present invention also discloses a smart bottle type recycling and identification method, wherein noise filtering and pixel expansion processing are performed simultaneously during the edge detection.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述處理模組對該邊緣影像進行像素侵蝕處理及像素膨脹處理取得一二值影像,並尋找二值影像中的外圍白點且記錄其四角座標點,再將設定像素以下的白點歸類為雜訊且排除,而其左上角為起始座標與右下角為最大座標所產生的外圍界定為所述定位框。The present invention also discloses a smart bottle type recycling identification method, wherein the processing module performs pixel erosion processing and pixel expansion processing on the edge image to obtain a binary image, and searches for peripheral white points in the binary image and records its four corners Then classify the white point below the set pixel as noise and eliminate it, and the periphery generated by the upper left corner as the starting coordinate and the lower right corner as the maximum coordinate is defined as the positioning frame.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述處理模組對第一基準區域與第二基準區域內之影像進行像素膨脹處理且計算其像素高度,若第一基準區域內之影像高度大於第二基準區域之影像高度時,該處理模組判斷該回收物品方向正確,若第一基準區域內之影像高度小於第二基準區域之影像高度時,該處理模組判斷該回收物品方向錯誤。The present invention also discloses a smart bottle type recycling identification method, wherein the processing module performs pixel expansion processing on the images in the first reference area and the second reference area and calculates the pixel height, if the image height in the first reference area If it is greater than the image height of the second reference area, the processing module judges that the orientation of the recycled item is correct; if the image height in the first reference area is smaller than the image height of the second reference area, the processing module determines that the orientation of the recycled item is wrong .
本發明另揭露一種智慧瓶型回收辨識方法,其中所述處理模組對第一基準區域與第二基準區域內之影像判斷該待回收物品方向錯誤時,退回所述待回收物品。The present invention also discloses a smart bottle type recycling identification method, wherein the processing module returns the item to be recycled when it judges that the direction of the item to be recycled is wrong based on the images in the first reference area and the second reference area.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述待回收物品方向正確時,該處理單元先對首處理影像以一輪廓搜索範圍搜尋該待回收物品之外包裝,並以一搜索條件搜索該外包裝之物品輪廓,再由物品輪廓界定所述局部框。The present invention also discloses a smart bottle-type recycling identification method, wherein when the direction of the item to be recycled is correct, the processing unit first searches the outer packaging of the item to be recycled with a contour search range for the first processed image, and searches for the outer packaging of the item to be recycled with a search condition The outline of the article in the outer package, and the partial frame is defined by the outline of the article.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述顏色設定條件為白色及綠色判斷該待回收物品進行回收,而顏色設定條件為黑色及藍色判斷該待回收物品退回。The present invention also discloses a smart bottle type recycling identification method, wherein the color setting conditions are white and green to determine that the item to be recycled is to be recycled, and the color setting conditions are black and blue to determine that the item to be recycled is returned.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述人工智慧模型辨識其待回收物品之材質及顏色。The present invention also discloses a smart bottle type recycling identification method, wherein the artificial intelligence model identifies the material and color of the items to be recycled.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述人工智慧模型辨識其待回收物品為瓶型且為PET材質時進行回收,若辨識其待回收物品為杯型且為PP材質或顏色為綠色時判斷該待回收物品退回。The present invention also discloses a smart bottle type recycling identification method, wherein the artificial intelligence model recognizes that the item to be recycled is bottle-shaped and made of PET material, and then recycles it. If it recognizes that the item to be recycled is cup-shaped and made of PP material or color is When it is green, it is judged that the item to be recycled is returned.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述處理模組取得所述首處理影像前,該處理模組先辨識該原始影像是否為寶特瓶,若為寶特瓶則取得所述首處理影像,若非為寶特瓶則退回該待回收物品。The present invention also discloses a smart bottle type recycling identification method, wherein before the processing module obtains the first processed image, the processing module first identifies whether the original image is a plastic bottle, and if it is a plastic bottle, then obtains the First process the image, and return the item to be recycled if it is not a PET bottle.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述回收機台更包括有一重量感測模組,該重量感測模組電性連接所述處理模組,且該重量感測模組係感測該待回收物品之重量並產生有一重量感測訊號傳遞至所述處理模組。The present invention also discloses a smart bottle recycling identification method, wherein the recycling machine further includes a weight sensing module, the weight sensing module is electrically connected to the processing module, and the weight sensing module is The weight of the item to be recycled is sensed and a weight sensing signal is generated and transmitted to the processing module.
本發明另揭露一種智慧瓶型回收辨識方法,其中所述回收機台更包括有一金屬感測模組,該金屬感測模組電性連接所述處理模組,且該金屬感測模組係感測該待回收物品並產生有一金屬感測訊號傳遞至所述處理模組。The present invention also discloses a smart bottle type recycling identification method, wherein the recycling machine further includes a metal sensing module, the metal sensing module is electrically connected to the processing module, and the metal sensing module is The object to be recycled is sensed and a metal sensing signal is generated and transmitted to the processing module.
本發明之上述目的及其結構與功能上的特性,將依據所附圖式之較佳實施例予以說明。The above-mentioned purpose of the present invention and its structural and functional characteristics will be described based on the preferred embodiments of the accompanying drawings.
在以下,針對本發明有關智慧瓶型回收辨識方法之構成及技術內容等,列舉各種適用的實例並配合參照隨文所附圖式而加以詳細地説明;然而,本發明當然不是限定於所列舉之該等的實施例、圖式或詳細說明內容而已。In the following, for the composition and technical content of the intelligent bottle type recycling identification method of the present invention, various applicable examples are listed and described in detail with reference to the accompanying drawings; however, the present invention is certainly not limited to the listed These embodiments, drawings, or detailed descriptions are just for reference.
再者,熟悉此項技術之業者亦當明瞭:所列舉之實施例與所附之圖式僅提供參考與說明之用,並非用來對本發明加以限制者;能夠基於該等記載而容易實施之修飾或變更而完成之發明,亦皆視為不脫離本發明之精神與意旨的範圍內,當然該等發明亦均包括在本發明之申請專利範圍。Furthermore, those who are familiar with this technology should also understand that the listed embodiments and accompanying drawings are only for reference and description, and are not used to limit the present invention; they can be easily implemented based on these records. Inventions completed through modification or alteration are also considered within the scope of not departing from the spirit and intent of the present invention, and of course such inventions are also included in the scope of the patent application of the present invention.
又,以下實施例所提到的方向用語,例如:「上」、「下」、「左」、「右」、「前」、「後」等,僅是參考附加圖示的方向。因此,使用的方向用語是用來說明,而並非用來限制本發明;再者,在下列各實施例中,相同或相似的元件將採用相同或相似的元件標號。In addition, the directional terms mentioned in the following embodiments, such as "upper", "lower", "left", "right", "front", "rear", etc., are only referring to the directions attached to the drawings. Therefore, the used directional terms are used to illustrate rather than limit the present invention; moreover, in the following embodiments, the same or similar components will use the same or similar component numbers.
請同時參閱第1圖及第2圖所示,係為本發明智慧瓶型回收辨識方法之流程圖及回收機台之硬體架構示意圖,其中本發明智慧瓶型回收辨識方法主要是應用於具有一回收機台1,該回收機台1設置有一影像擷取模組11及一處理模組12及至少一控制模組13及一人工智慧模型14及一重量感測模組15及一金屬感測模組16,該控制模組13設置有至少一背光組件131及至少一輸送組件132,其中該處理模組12訊號連接所述影像擷取模組11及該控制模組13及該人工智慧模型14及該重量感測模組15及金屬感測模組16,而智慧瓶型回收辨識方法的方法如下:Please refer to Figure 1 and Figure 2 at the same time, which are the flow chart of the smart bottle type recycling identification method of the present invention and the hardware structure diagram of the recycling machine, wherein the smart bottle type recycling identification method of the present invention is mainly applied to those with A recycling machine 1, the recycling machine 1 is provided with an
步驟S1:影像擷取模組擷取待回收物品之影像並取得有一原始影像;其中待回收物品投入該回收機台1之輸送組件132時,該控制模組13控制該輸送組件132動作且啟動所述背光模組,而該輸送組件132則將該待回收物品送至該背光模組前方,而該影像擷取模組11則擷取該待回收物品之影像並取得有一原始影像P1,如第3圖所示,而該原始影像P1則為該待回收物品及回收機台1內局部之影像,且該影像擷取模組11將所述原始影像P1傳送至所述處理模組12。Step S1: The image capture module captures the image of the article to be recycled and obtains an original image; when the article to be recycled is put into the conveying unit 132 of the recycling machine 1, the
步驟S2:處理模組接收所述原始影像且對原始影像之背景進行裁切並取得有一首處理影像;其中所述處理模組12接收所述原始影像P1後,該處理模組12會經由內部之儲存單元進行辨識,該儲存單元內則儲存有多筆寶特瓶或其它瓶型之外觀資料及材質資料,而該處理模組12則會先經由該儲存單元內之儲存資料與該原始影像P1進行比對辨識,若辨識該待回收物品非寶特瓶時,該處理模組12傳遞控制訊號至所述控制模組13,該控制模組13經由所述控制訊號控制該輸送組件132反轉且將待回收物品退回。Step S2: The processing module receives the original image and cuts the background of the original image to obtain a processed image; after the processing module 12 receives the original image P1, the processing module 12 will internally The storage unit is identified, and the storage unit stores the appearance data and material data of multiple PET bottles or other bottle types, and the processing module 12 will first pass the storage data in the storage unit and the original image P1 performs comparison and identification. If it is identified that the item to be recycled is not a plastic bottle, the processing module 12 transmits a control signal to the
而其該處理模組12比對辨識該待回收物品是否為寶特瓶之方式更包括可另外進行重量判斷之動作,其中該處理模組12之儲存單元內也儲存有各瓶型大小之相對重量範圍,因此,該處理模組12經由該儲存單元內之儲存資料與該原始影像P1進行比對辨識之同時,該處理模組12也會也會透過該瓶型之資料先得知該瓶型之相對重量範圍,而該重量感測模組15係感測該待回收物品之重量並產生有一重量感測訊號並傳送至該處理模組12,而該處理模組12接收該重量感測訊號並與該瓶型之相對重量範圍比對,若該重量感測訊號與該相對重量範圍比對相符時,則表示該待回收物品係為設定之寶特瓶且內無異物為空瓶並進行回收,反之,若該重量感測訊號與該相對重量範圍比對不相符時,則表示該待回收物品並非寶特瓶且內有異物非空瓶並將待回收物品退回並進行回收。The processing module 12 compares and identifies whether the article to be recycled is a plastic bottle, and includes an action of additional weight judgment, wherein the storage unit of the processing module 12 also stores the relative size of each bottle. Therefore, when the processing module 12 compares and identifies the original image P1 through the stored data in the storage unit, the processing module 12 will also first know the bottle through the data of the bottle type The relative weight range of the type, and the weight sensing module 15 senses the weight of the article to be recycled and generates a weight sensing signal and sends it to the processing module 12, and the processing module 12 receives the weight sensing The signal is compared with the relative weight range of the bottle type. If the weight sensing signal matches the relative weight range, it means that the item to be recycled is the set plastic bottle and there is no foreign matter in it. It is an empty bottle and Recycling, on the contrary, if the weight sensing signal does not match the relative weight range, it means that the item to be recycled is not a plastic bottle and there are foreign objects in it, not an empty bottle, and the item to be recycled will be returned and recycled.
又其中該處理模組12除了可透過外觀及材質與重量判斷待回收物品是否為寶特瓶外,更可透過金屬材質感測來辨識該待回收物品是否為寶特瓶或金屬瓶或內部有金屬異物,其中該金屬感測模組16係感測該感測該待回收物品是否具有金屬材質,若感測到該待回收物品具有金屬材質,則表示該待回收物係為金屬材質或內部有金屬異物,而該金屬感測模組16產生有金屬感測訊號傳遞至所述處理模組12,而該處理模組12接收所述金屬感測訊號後,便判定該待回收物品係為金屬材質或內部有金屬異物並將待回收物品退回。In addition, the processing module 12 can not only judge whether the article to be recycled is a plastic bottle through the appearance, material and weight, but also can identify whether the article to be recycled is a plastic bottle or a metal bottle or has a Metal foreign objects, wherein the
若該處理模組12辨識該待回收物品為寶特瓶且內部無異物時,該處理模組12對該原始影像P1多餘之背景影像進行裁切(ROI),其處理模組12先裁切多餘背景影像係可降低後續之計算量,而其處理模組12裁切多餘背景影像可取得有一首處理影像P2,如第4圖所示。If the processing module 12 recognizes that the article to be recycled is a plastic bottle and there is no foreign matter inside, the processing module 12 cuts (ROI) the redundant background image of the original image P1, and the processing module 12 cuts it first The redundant background image can reduce the amount of subsequent calculations, and the processing module 12 can obtain a processed image P2 by cutting the redundant background image, as shown in FIG. 4 .
步驟S3:再對首處理影像中之待回收物品進行邊緣檢測並取得一邊緣影像且於該邊緣影像外圍界定有一定位框,與將所述定位框形成於所述首處理影像上;該處理模組12取得所述首處理影像P2後,該處理模組12將該首處理影像P2中之待回收物品進行邊緣檢測,且於邊緣檢測時,於本實施例中係以該處理模組12先對所述首處理影像P2進行灰階處理,將原始RGB影像轉為灰階影像,如第5A圖所示,並該處理模組12將其灰階影像以高斯模糊7 7的核心大小進行影像模糊處理,如第5B圖所示,使像素強度均勻柔和,而後再將灰階處理及影像模糊處理後的影像以Canny邊緣檢測法進行邊緣檢測,而其係使用參數門檻24,以輸出帶有白色邊緣的二進制影像,如第5C圖所示,又其中,在邊緣檢測時,該處理模組12同時進行雜訊濾除處理及像素膨脹處理,其中該雜訊濾除處理主要是避免下方及四周雜訊白線條干擾,所以將其填補為0並進行濾除,如第5D圖所示,而於雜訊濾除後之白色像素再進行7 7的核心大小之膨脹處理,使其白色像素更加凸顯,如第5E圖所示,且該處理模組12透過其白色像素取得所述邊緣影像P3,而該處理模組12在取得邊緣影像P3後,該處理模組12為了將邊緣影像P3中之小雜點濾除且不破壞待回收物品的輪廓結構,該處理模組12便對其邊緣影像P3進行2 2的核心大小之像素侵蝕處理及進行20 20的核心大小之像素膨脹處理取得一二值影像,如第5F圖所示,且該處理模組12尋找二值影像中的外圍白點且記錄其四角座標點,且通過面積的計算,將1000像素以下的雜訊給排除,如第5G圖之黃框及第6圖所示,而圖中之藍色區域則為待回收物品之主要區塊,而該處理模組12於藍色區域擷取出後,將其左上角為起始座標與右下角為最大座標所產生的外圍界定為所述定位框F1,且該處理模組12將所述定位框F1形成於所述首處理影像P2上,如第6圖所示,而於本實施例中,係以處理模組12取得所述首處理影像P2後,該處理模組12將該首處理影像P2中之待回收物品進行邊緣檢測為主要實施方式,其中任何可以從首處理影像P2中取得邊緣影像P3的方式皆為本案保護範圍。 Step S3: Perform edge detection on the article to be recycled in the first processed image, obtain an edge image and define a positioning frame around the edge image, and form the positioning frame on the first processed image; the processing model After the group 12 obtains the first processed image P2, the processing module 12 performs edge detection on the articles to be recycled in the first processed image P2, and when detecting the edge, in this embodiment, the processing module 12 first Grayscale processing is performed on the first processed image P2, and the original RGB image is converted into a grayscale image, as shown in FIG. 5A, and the processing module 12 blurs the grayscale image by 7 The core size of 7 is used for image blur processing, as shown in Figure 5B, to make the pixel intensity uniform and soft, and then the image after gray scale processing and image blur processing is used for edge detection by Canny edge detection method, which uses the parameter threshold 24, to output a binary image with white edges, as shown in Figure 5C, and wherein, during edge detection, the processing module 12 simultaneously performs noise filtering processing and pixel expansion processing, wherein the noise filtering The main purpose of the processing is to avoid the interference of white lines of noise below and around, so it is filled with 0 and filtered, as shown in Figure 5D, and the white pixels after noise filtering are then processed by 7 The expansion processing of the kernel size of 7 makes its white pixels more prominent, as shown in Figure 5E, and the processing module 12 obtains the edge image P3 through its white pixels, and the processing module 12 obtains the edge image P3 Afterwards, in order to filter out the small noise points in the edge image P3 without destroying the outline structure of the article to be recycled, the processing module 12 performs 2 operations on the edge image P3. Pixel erosion processing with a core size of 2 and 20 A binary image is obtained by pixel expansion processing with a core size of 20, as shown in FIG. 5F, and the processing module 12 searches for peripheral white points in the binary image and records its four-corner coordinate points, and by calculating the area, the The noise below 1000 pixels is eliminated, as shown in the yellow frame of Figure 5G and Figure 6, and the blue area in the figure is the main block of the object to be recycled, and the processing module 12 is in the blue area After extraction, define the periphery generated by the upper left corner as the starting coordinate and the lower right corner as the maximum coordinate as the positioning frame F1, and the processing module 12 forms the positioning frame F1 on the first processed image P2 Above, as shown in Figure 6, in this embodiment, after the processing module 12 obtains the first processed image P2, the processing module 12 performs edge detection on the items to be recycled in the first processed image P2 It is the main implementation mode, and any method that can obtain the edge image P3 from the first processed image P2 is within the scope of protection of this case.
步驟S4:於該首處理影像之定位框內界定有一第一基準區域及一第二基準區域且對第一基準區域與第二基準區域內之影像判斷該待回收物品方向是否正確;該處理模組12將所述定位框F1形成於所述首處理影像P2上後,該處理模組12再於該首處理影像P2之定位框F1內界定有一第一基準區域B1及一第二基準區域B2,其中該第一基準區域B1係為該定位框F1內之左側,且於本實施例中係以定位框F1之左側往右移動130像素且將其間區域界定為第一基準區域B1,其中該第二基準區域B2係為該定位框F1內之右側,且其是定位框F1之右側往左移動130像素且將其間區域界定為第二基準區域B2,而後再裁切第一基準區域B1及第二基準區域B2外之區域,且該處理模組12對第一基準區域B1與第二基準區域B2內之影像進行像素膨脹處理且計算其像素高度,若第一基準區域B1內之影像高度大於第二基準區域B2之影像高度時,該處理模組12判斷該回收物品方向正確,若第一基準區域B1內之影像高度小於第二基準區域B2之影像高度時,該處理模組12判斷該回收物品方向錯誤,且退回所述待回收物品,如第7A及7B圖所示,且於本實施例中,係以第一基準區域B1與第二基準區域B2間之高度比對來判斷該待回收物品方向是否正確為主要實施方式,其中任何可以判斷該待回收物品方向方式皆為本案保護範圍。Step S4: Defining a first reference area and a second reference area in the positioning frame of the first processed image and judging whether the direction of the object to be recycled is correct for the images in the first reference area and the second reference area; the processing model After group 12 forms the positioning frame F1 on the first processed image P2, the processing module 12 defines a first reference area B1 and a second reference area B2 within the positioning frame F1 of the first processed image P2 , wherein the first reference area B1 is the left side of the positioning frame F1, and in this embodiment, the left side of the positioning frame F1 is moved to the right by 130 pixels and the area in between is defined as the first reference area B1, wherein the The second reference area B2 is the right side of the positioning frame F1, and it moves the right side of the positioning frame F1 to the left by 130 pixels and defines the area in between as the second reference area B2, and then cuts the first reference area B1 and The area outside the second reference area B2, and the processing module 12 performs pixel expansion processing on the images in the first reference area B1 and the second reference area B2 and calculates the pixel height, if the height of the image in the first reference area B1 If it is greater than the image height of the second reference area B2, the processing module 12 judges that the direction of the recycled article is correct; if the image height in the first reference area B1 is smaller than the image height of the second reference area B2, the processing module 12 judges The recycling item is in the wrong direction, and the item to be recycled is returned, as shown in Figures 7A and 7B, and in this embodiment, it is judged based on the height comparison between the first reference area B1 and the second reference area B2 Whether the direction of the article to be recycled is correct is the main implementation method, and any method that can determine the direction of the article to be recycled is within the scope of protection of this case.
步驟S5:待回收物品方向正確時,該處理單元再對首處理影像中之定位框內界定有至少一局部框,由局部框內之顏色判定該待回收物品之顏色且以一顏色設定條件判斷該待回收物品是否進行回收;其中處理模組12判斷該待回收物品方向正確時,該處理模組12再對首處理影像P2中之定位框F1內界定有至少一局部框F2,而該局部框F2界定前,該處理單元先對首處理影像P2以一輪廓搜索範圍搜尋該待回收物品之外包裝,其中該輪廓搜索範圍最小值於本實施例中係以界定為定位框F1X軸最小值座標+定位框F1寬度/5,及輪廓搜索範圍最大值係以界定為定位框F1X軸最大值座標+定位框F1寬度/4,但不因此為限,所以該處理單元經由所述輪廓搜索範圍來找出該待回收物品可能所形成有一輪廓範圍,該處理單元再以一搜索條件搜索該待回收物品之輪廓,其中所述搜索條件於本實施例中之首要條件為該輪廓長度大於10pixel及寬度大於10pixel之條件下,該處理單元才會將此標的設定為待回收物品之輪廓,但不因此為限,而於輪廓設定後,再進行搜索條件來抓取其輪廓位置,而輪廓位置之搜索條件為輪廓範圍X軸最小值 定位框F1X軸最小值-50pixel、輪廓範圍X軸最大值 定位框F1X軸最大值+60pixel、輪廓範圍Y軸最小值 定位框F1Y軸最小值-50pixel、輪廓範圍Y軸最大值 定位框F1Y軸最大值+60pixel,所以該處理單元透過所述搜索條件取得該待回收物品之物品輪廓,而其物品輪廓可為底部凹槽輪廓或包裝輪廓或頭部輪廓,而該處理單元便可由該影像判斷出該物品輪廓數量。 Step S5: When the direction of the article to be recycled is correct, the processing unit then defines at least one partial frame in the positioning frame in the first processed image, determines the color of the article to be recycled based on the color in the partial frame, and sets a color to determine the condition Whether the article to be recycled is to be recycled or not; when the processing module 12 judges that the direction of the article to be recycled is correct, the processing module 12 defines at least one local frame F2 in the positioning frame F1 in the first processed image P2, and the local Before the frame F2 is defined, the processing unit searches the first processed image P2 with a contour search range to search for the outer packaging of the article to be recycled, wherein the minimum value of the contour search range is defined as the minimum value of the X-axis of the positioning frame F1 in this embodiment Coordinates + positioning frame F1 width/5, and the maximum value of the contour search range are defined as the positioning frame F1 X axis maximum coordinates + positioning frame F1 width/4, but not limited thereto, so the processing unit passes through the contour search range To find out the possible outline range of the article to be recycled, the processing unit searches the outline of the article to be recycled with a search condition, wherein the first condition of the search condition in this embodiment is that the length of the outline is greater than 10pixel and Only when the width is greater than 10pixels, the processing unit will set the target as the outline of the object to be recycled, but not limited to this. After the outline is set, search conditions are used to capture the outline position, and the outline position The search condition is the minimum value of the contour range X axis The minimum value of the positioning frame F1 X-axis -50pixel, the maximum value of the contour range X-axis Positioning frame F1 X-axis maximum value +60pixel, outline range Y-axis minimum value The minimum value of the F1 Y axis of the positioning frame -50pixel, the maximum value of the Y axis of the outline range The maximum value of the F1Y axis of the positioning frame is +60pixel, so the processing unit obtains the item outline of the item to be recycled through the search conditions, and the item outline can be the bottom groove outline or the package outline or the head outline, and the processing unit then The outline quantity of the object can be judged from the image.
若物品輪廓數量 2時,其物品輪廓由左至右之順序則可為包裝輪廓及頭部輪廓、包裝輪廓及包裝輪廓及頭部輪廓、底部凹槽輪廓及包裝輪廓及頭部輪廓,並於輪廓數量取得後,該處理單元會再計算位置偏差值來產生至少一局部框F2,而最左邊的局部框F2其位置偏差值可舉例為在100pixel,位置偏差值(輪廓範圍X軸之位置-定位框F1X軸最小值) 50pixel之狀況下產生有局部框F2,並該處理單元計算該局部框F2所要產生之位移量,如100pixel/4=移動25pixel,而局部框F2之位置與大小則為:局部框F2X軸最小值=定位框F1X軸最小值+25pixel、與局部框F2X軸最大值=輪廓範圍X軸最小值-25pixel。 If the item outline quantity 2, the order from left to right of the outline of the article can be the outline of the package and the head, the outline of the package and the outline of the head, the outline of the bottom groove, the outline of the package and the head, and after the number of outlines is obtained , the processing unit will then calculate the position deviation value to generate at least one local frame F2, and the position deviation value of the leftmost local frame F2 can be, for example, at 100 pixels, the position deviation value (the position of the X-axis of the contour range-the position of the F1 X-axis of the positioning frame min) In the case of 50 pixels, a local frame F2 is generated, and the processing unit calculates the displacement amount of the local frame F2, such as 100pixel/4 = moving 25 pixels, and the position and size of the local frame F2 are: the minimum value of the X axis of the local frame F2 = The minimum value of the F1 X axis of the positioning frame + 25pixel, and the maximum value of the F2 X axis of the local frame = The minimum value of the X axis of the outline range - 25pixel.
而最右邊的局部框F2其位置偏差值則為該位置X軸最小值-上一個局部框F2X軸最大值,並該處理單元計算該局部框F2所要產生之位移量,如100pixel/4=移動25pixel,且其中也可具有一設定條件,若該位置X軸最小值不在輪廓範圍內時,則該設定位置偏差值在 8pixel內時,則產生有局部框F2;若位置X軸最小值在輪廓範圍內時,則該設定位置偏差值在 90pixel內時,則產生有局部框F2,而局部框F2之位置與大小則為:局部框F2X軸最小值=上一個局部框F2X軸最大值+25pixel、與局部框F2X軸最大值=輪廓範圍X軸最小值-25pixel,如第8A圖所示。 The position deviation value of the rightmost local frame F2 is the minimum value of the X axis of this position - the maximum value of the X axis of the previous local frame F2, and the processing unit calculates the displacement to be generated by the local frame F2, such as 100pixel/4=move 25pixel, and there can also be a setting condition, if the minimum value of the X-axis of the position is not within the contour range, then the set position deviation value is within If it is within 8pixel, a local frame F2 will be generated; if the minimum value of the position X axis is within the contour range, the set position deviation value is within When within 90 pixels, there is a local frame F2, and the position and size of the local frame F2 are: the minimum value of the F2X axis of the local frame = the maximum value of the F2X axis of the previous local frame + 25pixel, and the maximum value of the F2X axis of the local frame = the outline range The minimum value of the X axis is -25pixel, as shown in Figure 8A.
若物品輪廓數量只有一個時,位置偏差值=定位框F1X軸最小值-輪廓範圍X軸最大值,且位置偏差值(輪廓範圍X軸之位置-定位框F1X軸最小值) 50pixel之狀況下產生有局部框F2,並該處理單元計算該局部框F2所要產生之位移量,如100pixel/4=移動25pixel,而局部框F2之位置與大小則為:局部框F2X軸最小值=輪廓範圍X軸最小值+25pixel、與局部框F2X軸最大值=定位框F1X軸最大值-25pixel,若位置偏差值(輪廓範圍X軸之位置-定位框F1X軸最小值) 50pixel之狀況下,則計算其整個待回收物品之有局部框F2,而局部框F2之位置與大小則為:局部框F2X軸最小值=定位框F1X軸最小值、局部框F2X軸最大值=定位框F1X軸最大值,如第8B圖所示,而其中局部框F2之限制條件與計算方式並不因此為限,僅要在定位框F1內界定有局部框F2皆為本案保護範圍。 If there is only one outline of the item, the position deviation value = the minimum value of the F1X axis of the positioning frame - the maximum value of the X axis of the contour range, and the position deviation value (the position of the X axis of the contour range - the minimum value of the F1X axis of the positioning frame) In the case of 50 pixels, a local frame F2 is generated, and the processing unit calculates the displacement amount of the local frame F2, such as 100pixel/4 = moving 25 pixels, and the position and size of the local frame F2 are: the minimum value of the X axis of the local frame F2 =Minimum value of the X-axis of the contour range+25pixel, and the maximum value of the F2X-axis of the local frame=The maximum value of the F1X-axis of the positioning frame-25pixel, if the position deviation value (the position of the X-axis of the contour range-the minimum value of the F1X-axis of the positioning frame) In the case of 50 pixels, calculate the local frame F2 of the entire item to be recycled, and the position and size of the local frame F2 are: the minimum value of the X axis of the local frame F2 = the minimum value of the X axis of the positioning frame F1, the maximum value of the X axis of the local frame F2 = The maximum value of the X-axis of the positioning frame F1 is shown in Figure 8B, and the constraints and calculation methods of the local frame F2 are not limited thereto. As long as the local frame F2 is defined within the positioning frame F1, it is within the scope of protection of this case.
而該處理單元界定有局部框F2後,該處理單元再由由局部框F2內之顏色判定該待回收物品之顏色且以一顏色設定條件判斷該待回收物品是否進行回收,例如其中顏色設定條件為白色及綠色判斷該待回收物品進行回收,而顏色設定條件為黑色及藍色判斷該待回收物品退回,而其中顏色條件設定並不因此為限,該顏色設定條件中所設定之顏色係可依照使用者之需求進行設定,例如顏色設定條件為黑色及藍色判斷該待回收物品進行回收,而顏色設定條件為白色及綠色判斷該待回收物品退回。And after the processing unit defines a partial frame F2, the processing unit judges the color of the article to be recycled by the color in the partial frame F2 and judges whether the article to be recycled is recycled by a color setting condition, for example, the color setting condition The item to be recycled is judged as white and green for recycling, and the color setting condition is black and blue to judge that the item to be recycled is returned, and the setting of the color condition is not limited to this, the color set in the color setting condition can be Set according to the needs of users. For example, if the color setting condition is black and blue, it is judged that the item to be recycled is recycled, while the color setting condition is white and green, it is judged that the item to be recycled is returned.
步驟S6:待回收物品顏色符合顏色條件設定時,再以一人工智慧模型對該待回收物品進行辨識。Step S6: When the color of the item to be recycled meets the color condition setting, an artificial intelligence model is used to identify the item to be recycled.
其中所述處理單元判斷該待回收物品顏色符合顏色條件設定時,該處理單元則將所述首處理影像P2傳送至人工智慧模型14,該人工智慧模型14係以yolo類神經模型架構,其主要是學習有PET類型瓶型、PP類型瓶型及綠色類型,而該人工智慧模型14接收有所述首處理影像P2後,該人工智慧模型14對該首處理影像P2進行辨識,以對該待回收物品進行最終辨識,人工智慧模型14辨識其待回收物品為瓶型且為PET材質時進行回收,若辨識其待回收物品為杯型且為PP材質或顏色為綠色時判斷該待回收物品退回,如第9圖所示。When the processing unit determines that the color of the item to be recycled meets the color condition setting, the processing unit then transmits the first processed image P2 to the artificial intelligence model 14. The artificial intelligence model 14 is based on a yolo-like neural model architecture, and its main After learning the PET type bottle type, PP type bottle type and green type, and the artificial intelligence model 14 receives the first processed image P2, the artificial intelligence model 14 recognizes the first processed image P2, so as to Recycled items are finally identified. The artificial intelligence model 14 recognizes that the item to be recycled is bottle-shaped and made of PET material, and then recycles it. If it recognizes that the item to be recycled is cup-shaped and made of PP material or the color is green, it is judged that the item to be recycled is returned. , as shown in Figure 9.
因此,本實施例中,該智慧瓶型回收辨識方法所回收之待回收物品之寶特瓶材質需要為透明之PET材質且內部不能具有液體或異物,但寶特瓶之瓶蓋與包膜並在透光度之限制之中,並且經踩踏過之寶特瓶也是可回收之範疇,另外,本實施例中係以透明之PET材質為主要實施方式,但不因此為限,主要是可依照回收物品之種類進行處理模組12之設定,例如也可設定為僅回收PP材質的瓶罐,其主要就是經由處理模組12設定材質光譜便可變換其回收之材質,藉此達到可精準判斷回收物以節省人力與時間耗費之功效者。Therefore, in this embodiment, the plastic bottle material of the items to be recycled recovered by the smart bottle type recycling identification method needs to be a transparent PET material and there must be no liquid or foreign matter inside, but the bottle cap and the coating of the plastic bottle are not the same. Under the limitation of light transmittance, plastic bottles that have been stepped on are also recyclable. In addition, in this embodiment, transparent PET material is used as the main implementation method, but it is not limited to this, mainly according to The processing module 12 sets the type of recycled items. For example, it can also be set to only recycle PP bottles and cans. The main reason is that the recycled materials can be changed by setting the material spectrum through the processing module 12, so as to achieve accurate judgment. Recycling can save manpower and time consumption.
以上已將本發明做一詳細說明,惟以上所述者,僅為本發明之一較佳實施例而已,當不能限定本發明實施之範圍,即凡依本發明申請範圍所作之均等變化與修飾等,皆應仍屬本發明之專利涵蓋範圍。The present invention has been described in detail above, but the above is only a preferred embodiment of the present invention, and should not limit the scope of the present invention, that is, all equivalent changes and modifications made according to the application scope of the present invention etc., should still fall within the scope of the patent coverage of the present invention.
1:回收機台 11:影像擷取模組 12:處理模組 13:控制模組 131:背光組件 132:輸送組件 14:人工智慧模型 15:重量感測模組 16:金屬感測模組 P1:原始影像 P2:首處理影像 P3:邊緣影像 F1:定位框 B1:第一基準區域 B2:第二基準區域 F2:局部框 S1~S6:步驟 1: Recycling machine 11: Image capture module 12: Processing module 13: Control module 131:Backlight assembly 132: Conveying components 14: Artificial intelligence model 15: Weight sensing module 16: Metal sensing module P1: Original image P2: First processing image P3: Edge image F1: positioning frame B1: the first reference area B2: Second base area F2: Local frame S1~S6: steps
第1圖係為本發明智慧瓶型回收辨識方法之流程圖。 第2圖係為本發明回收機台之硬體架構示意圖。 第3圖係為本發明原始影像示意圖。 第4圖係為本發明首處理影像示意圖。 第5A~5G圖係為本發明邊緣影像產生示意圖。 第6圖係為本發明定位框產生示意圖。 第7A~7B圖係為本發明方向判斷示意圖。 第8A~8B圖係為本發明局部框產生示意圖。 第9圖係為本發明人工智慧模型辨識示意圖。 Fig. 1 is a flow chart of the intelligent bottle recycling identification method of the present invention. Figure 2 is a schematic diagram of the hardware structure of the recycling machine of the present invention. Figure 3 is a schematic diagram of the original image of the present invention. Fig. 4 is a schematic diagram of the first processed image of the present invention. Figures 5A-5G are schematic diagrams of edge image generation according to the present invention. Figure 6 is a schematic diagram of the creation of the positioning frame of the present invention. Figures 7A-7B are schematic diagrams of the direction judgment of the present invention. Figures 8A-8B are schematic diagrams of partial frame generation in the present invention. Fig. 9 is a schematic diagram of artificial intelligence model identification of the present invention.
S1~S6:步驟 S1~S6: steps
Claims (12)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW111139935A TWI805513B (en) | 2022-10-21 | 2022-10-21 | Intelligent bottle recycling identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW111139935A TWI805513B (en) | 2022-10-21 | 2022-10-21 | Intelligent bottle recycling identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
TWI805513B true TWI805513B (en) | 2023-06-11 |
TW202418120A TW202418120A (en) | 2024-05-01 |
Family
ID=87803492
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW111139935A TWI805513B (en) | 2022-10-21 | 2022-10-21 | Intelligent bottle recycling identification method |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI805513B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361681A (en) * | 2014-11-20 | 2015-02-18 | 北京盈创高科新技术发展有限公司 | Beverage bottle recognition and recovery method and device for intelligent recovery machine |
TWM580466U (en) * | 2019-03-12 | 2019-07-11 | 凡立橙股份有限公司 | Structure of bottle recycling system |
CN110969267A (en) * | 2019-12-18 | 2020-04-07 | 恒大智慧科技有限公司 | Community garbage processing method, community server and computer readable storage medium |
CN111715674A (en) * | 2020-07-03 | 2020-09-29 | 广东隽诺环保工程技术有限公司 | Intelligent sorting process and system for garbage classification high-value recyclable materials |
US20210371196A1 (en) * | 2016-07-15 | 2021-12-02 | Cleanrobotics Technologies, Inc. | Automatic sorting of waste |
-
2022
- 2022-10-21 TW TW111139935A patent/TWI805513B/en active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361681A (en) * | 2014-11-20 | 2015-02-18 | 北京盈创高科新技术发展有限公司 | Beverage bottle recognition and recovery method and device for intelligent recovery machine |
US20210371196A1 (en) * | 2016-07-15 | 2021-12-02 | Cleanrobotics Technologies, Inc. | Automatic sorting of waste |
TWM580466U (en) * | 2019-03-12 | 2019-07-11 | 凡立橙股份有限公司 | Structure of bottle recycling system |
CN110969267A (en) * | 2019-12-18 | 2020-04-07 | 恒大智慧科技有限公司 | Community garbage processing method, community server and computer readable storage medium |
CN111715674A (en) * | 2020-07-03 | 2020-09-29 | 广东隽诺环保工程技术有限公司 | Intelligent sorting process and system for garbage classification high-value recyclable materials |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111896556B (en) | Glass bottle bottom defect detection method and system based on machine vision | |
CA3215397A1 (en) | Methods and arrangements to aid recycling | |
CN106370659A (en) | Visual inspection method for bottle packaging quality | |
CN110116415A (en) | A kind of Bottle & Can class rubbish identification sorting machine people based on deep learning | |
CN106548111B (en) | A kind of two dimensional code open defect detection method and device | |
CN109815856A (en) | Status indication method, system and the computer readable storage medium of target vehicle | |
EP3865222A1 (en) | A method for sorting consumer packaging objects travelling on a conveyor belt | |
CN113034488B (en) | Visual inspection method for ink-jet printed matter | |
CN111652214B (en) | Garbage bottle sorting method based on deep learning | |
CN114549504A (en) | Appearance quality detection method based on machine vision | |
CN113191358B (en) | Metal part surface text detection method and system | |
CN112557412A (en) | Automatic code-spraying printing defect detection system and detection method | |
TWI805513B (en) | Intelligent bottle recycling identification method | |
CN113724322A (en) | Cargo pallet positioning method and system for unmanned forklift | |
CN114049556A (en) | Garbage classification method integrating SVM (support vector machine) and target detection algorithm | |
TW202418120A (en) | Smart bottle type recycling identification method | |
Huang et al. | A real-time algorithm for aluminum surface defect extraction on non-uniform image from CCD camera | |
WO2024015385A1 (en) | Methods and arrangements to utilize end-of-life data generated during recycling and waste sortation for counterfeit deterrence and other actions | |
CN105740768A (en) | Unmanned forklift device based on combination of global and local features | |
Wang et al. | The two-step method of pavement pothole and raveling detection and segmentation based on deep learning | |
CN111191631A (en) | Video monitoring supervision area multi-device linkage method based on image domain template matching | |
Yani et al. | Development of identification system of cans and bottle | |
CN102073868A (en) | Digital image closed contour chain-based image area identification method | |
CN114986520A (en) | Four-axis parallel robot sorting system and method based on machine vision | |
Li et al. | Detection of containerized containers based on computer vision |