TWI784740B - Poultry health monitoring system and method thereof - Google Patents

Poultry health monitoring system and method thereof Download PDF

Info

Publication number
TWI784740B
TWI784740B TW110137175A TW110137175A TWI784740B TW I784740 B TWI784740 B TW I784740B TW 110137175 A TW110137175 A TW 110137175A TW 110137175 A TW110137175 A TW 110137175A TW I784740 B TWI784740 B TW I784740B
Authority
TW
Taiwan
Prior art keywords
poultry
image
module
weight
monitoring system
Prior art date
Application number
TW110137175A
Other languages
Chinese (zh)
Other versions
TW202315518A (en
Inventor
張光甫
黃醴萬
張家榕
Original Assignee
智逐科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 智逐科技股份有限公司 filed Critical 智逐科技股份有限公司
Priority to TW110137175A priority Critical patent/TWI784740B/en
Application granted granted Critical
Publication of TWI784740B publication Critical patent/TWI784740B/en
Publication of TW202315518A publication Critical patent/TW202315518A/en

Links

Images

Abstract

A poultry health monitoring system includes: a cloud module, a learning calibration module, and a monitoring module. The learning calibration module is used to sense a weight of at least one poultry carried and a first poultry image to analyze the number of poultry in the first poultry image. And the learning calibration module generates a poultry image characteristic and a weight formula. The monitoring module is used to generate a second poultry image including at least one poultry. The cloud module obtains a unit weight, and an activity value of each poultry based on the second poultry image. The present disclosure further includes a method for monitoring a health states of poultry.

Description

家禽健康監測系統及其方法 Poultry health monitoring system and method thereof

本發明係有關一種健康監測系統及其方法,尤指針對家禽且可自動化校正與監測以節約人力的一種家禽健康監測系統及其方法。 The invention relates to a health monitoring system and method thereof, in particular to a poultry health monitoring system and method that can be automatically corrected and monitored to save manpower.

家禽中的雞隻與家畜中的豬隻一直是我們生活飲食中蛋白質的主要攝取來源,不僅營養價值高,也是許多加工食品的主要原料。近年來,臺灣的雞肉商品經濟產值達391億新台幣,佔臺灣畜牧業總產值23.92%,是國內重要的農產品。雞隻健康狀態與雞隻的飲食行為常有緊密關聯,目前禽畜舍的飲食行為偵測則多以人工方式為主;然而禽畜舍雞隻數量龐大,以傳統管理方式相當耗時、勞力密集且仰賴飼主經驗,容易在成本控制以及品質控管上出現問題。 Chickens in poultry and pigs in livestock have always been the main sources of protein intake in our daily diet. They not only have high nutritional value, but are also the main raw materials for many processed foods. In recent years, the economic output value of Taiwan's chicken commodity has reached NT$39.1 billion, accounting for 23.92% of the total output value of Taiwan's animal husbandry industry. It is an important domestic agricultural product. The health status of chickens is often closely related to the eating behavior of chickens. At present, the detection of eating behaviors in poultry houses is mostly manual. However, the number of chickens in poultry houses is huge, and traditional management methods are quite time-consuming and labor Intensive and dependent on the experience of the owner, it is easy to have problems in cost control and quality control.

然而,在家禽養殖的期間,往往受限於監控裝置的數量、範圍或反應速度等因素,而造成來不及挽回的損失。尤其是在熱帶和亞熱帶地區,由於夏季氣候炎熱,熱緊迫(heat stress)成為家禽養殖業者最具挑戰性的問題之一。熱緊迫降低了家禽的生長速度且對雞蛋質量產生不利影響,甚至與家禽的突然大量死亡有關。早期的養殖經驗發現,熱緊迫是穩定家禽成長和雞蛋品質的關鍵,通常評估熱緊迫是使用溫濕度指數 (temperature humidity index,THI)估計,也就是同時量測溫度以及濕度來進行評估。然而,溫溼度指數是間接指標,而熱緊迫的標準也可能因雞隻品種和雞隻的膳食及飲水供應而有差異,容易造成熱緊迫或生長狀況之評估錯誤,而致使家禽養殖上的損失。 However, during the period of poultry breeding, it is often limited by factors such as the number, range, or response speed of monitoring devices, resulting in irreversible losses. Especially in the tropics and subtropics, heat stress is one of the most challenging issues for poultry farmers due to the hot summer climate. Heat stress reduces the growth rate of poultry and adversely affects egg quality, and has even been associated with sudden mass mortality of poultry. Early farming experience found that heat stress is the key to stabilizing poultry growth and egg quality, and heat stress is usually assessed using the temperature and humidity index (temperature humidity index, THI) estimation, which is to measure the temperature and humidity at the same time for evaluation. However, the temperature and humidity index is an indirect indicator, and the standard of heat stress may also vary depending on the breed of chicken and the food and drinking water supply of the chicken, which may easily lead to errors in the assessment of heat stress or growth conditions, resulting in losses in poultry farming .

且進一步而言,對於一般家禽飼養業者來說,其深知家禽之健康狀況與其各體重量以及活動力息息相關,若於家禽之飼養過程以及成長過程期間,各體重量不足、或給予的活動空間或活動時間不足時,則會嚴重影響飼養家禽之健康狀況。傳統上若要解決這個問題,通常需要投入大量人力來進行各家禽個體之體重量測、現場觀察評估各家禽之健康狀況、以及人力紀錄活動力以及活動時間等等行為。如此一來不僅往往花費大量人力與時間成本,且處理速度亦無法提升,造成家禽飼養業者之飼養維護成本難以降低、以及監控效率不彰之技術問題。 And further, for general poultry farmers, they are well aware that the health status of poultry is closely related to their weight and activity. Or when the activity time is insufficient, it will seriously affect the health of the poultry. Traditionally, to solve this problem, it usually requires a lot of manpower to measure the weight of each poultry individual, observe and evaluate the health status of each poultry on the spot, and manually record the activity and activity time and other behaviors. In this way, not only often a lot of manpower and time costs are spent, but also the processing speed cannot be improved, resulting in the technical problems that it is difficult for poultry farmers to reduce the cost of feeding and maintenance, and the monitoring efficiency is not good.

為此,如何設計出一種家禽健康監測系統及其方法,來解決前述的技術問題,乃為本案發明人所研究的重要課題。 For this reason, how to design a poultry health monitoring system and method thereof to solve the aforementioned technical problems is an important topic studied by the inventor of this case.

本發明之目的在於提供一種家禽健康監測系統,可以解決現有技術之飼養維護成本難以降低、以及監控效率不彰之技術問題,達到低維護成本、可快速反應且全時監控之目的。 The purpose of the present invention is to provide a poultry health monitoring system, which can solve the technical problems of difficulty in reducing feeding and maintenance costs and poor monitoring efficiency in the prior art, and achieve the goals of low maintenance costs, quick response and full-time monitoring.

為了達到前述目的,本發明所提出的所述家禽健康監測系統包括:雲端模組、運算核心、學習校正模組以及監控模組。其中,雲端模組用以儲存至少一家禽影像特徵、以及對應各家禽影像特徵之一影像重 量關係式。運算核心耦接雲端模組,該運算核心接收該重量值以及該第一家禽影像,以分析該第一家禽影像中之該至少一家禽的數量,且產生該至少一家禽影像特徵以及該影像重量關係式,且該影像重量關係式包括影像特徵值與重量之相對關係。學習校正模組耦接該運算核心以及該雲端模組,且包括一秤重結構、一第一攝影機;其中,該秤重結構用以感測所承載之至少一家禽的重量值;該第一攝影機配置於該秤重結構中,且用以產生該秤重結構所承載之該至少一家禽的一第一家禽影像。監控模組耦接該雲端模組,且包括一第二攝影機;其中,該第二攝影機用以產生包括至少一該家禽的一第二家禽影像。其中,該雲端模組依據該第二家禽影像、該至少一家禽影像特徵、以及該影像重量關係式,而獲得各該家禽的一單位重量。 In order to achieve the aforementioned purpose, the poultry health monitoring system proposed by the present invention includes: a cloud module, a computing core, a learning correction module and a monitoring module. Wherein, the cloud module is used to store at least one poultry image feature, and an image weight corresponding to each poultry image feature. quantitative relationship. The computing core is coupled to the cloud module, the computing core receives the weight value and the first poultry image, analyzes the quantity of the at least one poultry in the first poultry image, and generates the at least one poultry image feature and the image weight Relational expression, and the image weight relational expression includes the relative relationship between image feature value and weight. The learning correction module is coupled to the computing core and the cloud module, and includes a weighing structure and a first camera; wherein, the weighing structure is used to sense the weight value of at least poultry carried; the first The camera is arranged in the weighing structure and is used to generate a first poultry image of the at least one poultry carried by the weighing structure. The monitoring module is coupled to the cloud module and includes a second camera; wherein, the second camera is used to generate a second poultry image including at least one poultry. Wherein, the cloud module obtains a unit weight of each poultry according to the second poultry image, the at least one poultry image feature, and the image weight relational expression.

進一步而言,本發明所述之家禽健康監測系統更包括預警分析模組,所述預警分析模組耦接所述雲端模組,且所述預警分析模組依據所述單位重量、活動力之至少一者,以輸出統計報告以及警示訊息之至少一者。 Furthermore, the poultry health monitoring system of the present invention further includes an early warning analysis module, the early warning analysis module is coupled to the cloud module, and the early warning analysis module is based on the unit weight, activity At least one, to output at least one of statistical reports and warning messages.

進一步而言,本發明所述之家禽健康監測系統更包括行動通訊平台,所述行動通訊平台無線地耦接所述預警分析模組,且接收所述統計報告以及所述警示訊息之至少一者。 Furthermore, the poultry health monitoring system of the present invention further includes a mobile communication platform, the mobile communication platform is wirelessly coupled to the early warning analysis module, and receives at least one of the statistical report and the warning message .

進一步而言,本發明所述之家禽健康監測系統中,所述行動通訊平台包括工作站、伺服器、桌上型電腦、筆記型電腦、平板電腦、個人數位助理或智慧型手機的其中一者。 Furthermore, in the poultry health monitoring system of the present invention, the mobile communication platform includes one of a workstation, a server, a desktop computer, a notebook computer, a tablet computer, a personal digital assistant or a smart phone.

進一步而言,本發明所述之家禽健康監測系統中,該運算核心為利用一物件偵測演算工具作為該運算核心辨識目標物的深度學習架構,所述物件偵測演算工具為深度學習或影像處理方式,該雲端模組藉由該至少一卷積層與該至少一池化層以比對該第二家禽影像是否符合各該家禽影像特徵,繼而獲得單位重量、活動力數值。 Furthermore, in the poultry health monitoring system described in the present invention, the calculation core is a deep learning framework that uses an object detection calculation tool as the calculation core to identify the target object, and the object detection calculation tool is deep learning or image In the processing method, the cloud module uses the at least one convolutional layer and the at least one pooling layer to compare whether the second poultry image conforms to the characteristics of each poultry image, and then obtains unit weight and activity value.

進一步而言,本發明所述之家禽健康監測系統中,該雲端模組包括一伺服器以及一雲端資料庫;其中該伺服器用以獲得該單位重量、該活動力數值之至少一者;該雲端資料庫耦接該伺服器,且用以儲存該至少一家禽影像特徵、該影像重量關係式、該單位重量、活動力之至少一者 Further, in the poultry health monitoring system of the present invention, the cloud module includes a server and a cloud database; wherein the server is used to obtain at least one of the unit weight and the activity value; the The cloud database is coupled to the server, and is used to store at least one of the at least one poultry image feature, the image weight relation, the unit weight, and the activity force

進一步而言,本發明所述之家禽健康監測系統中,所述伺服器藉由窄頻物聯網(narrowband internet of things,NB-Iot)、LoRa WAN、LTE以及Wi-Fi之其中一者耦接所述雲端資料庫。 Further, in the poultry health monitoring system of the present invention, the server is coupled by one of narrowband internet of things (NB-Iot), LoRa WAN, LTE and Wi-Fi The cloud database.

進一步而言,本發明所述之家禽健康監測系統中,所述秤重結構中包括秤重平台以及中間平台。其中所述秤重平台用以承載至少一所述家禽,所述中間平台配置於所述秤重平台之上,且所述第一攝影機配置於所述中間平台之下。 Furthermore, in the poultry health monitoring system of the present invention, the weighing structure includes a weighing platform and an intermediate platform. Wherein the weighing platform is used to carry at least one poultry, the middle platform is arranged on the weighing platform, and the first camera is arranged under the middle platform.

進一步而言,本發明所述之家禽健康監測系統中,所述秤重平台藉由至少兩個柱體耦接所述中間平台。 Furthermore, in the poultry health monitoring system of the present invention, the weighing platform is coupled to the intermediate platform by at least two columns.

在使用本發明所述之家禽健康監測系統及其方法時,所述家禽健康監測系統之所述雲端模組可預先儲存有重量判斷資料庫。並且,首先由所述學習校正模組先進行人工智慧(artificial intelligence,AI)模型 之機器學習(machine learning,ML)程序,所述學習校正模組先以所述秤重結構感測所承載之至少一家禽的重量值,且以所述第一攝影機產生所述秤重結構所承載之所述至少一家禽的所述第一家禽影像。最後學習校正模組再以所述運算核心接收所述重量值以及所述第一家禽影像,以分析所述第一家禽影像中之所述至少一家禽的數量,且產生所述至少一家禽影像特徵以及所述影像重量關係式以儲存於所述雲端模組,藉此完成所述機器學習程序。使得所述雲端模組儲存至少一家禽影像特徵、影像重量關係式。相對於前述步驟可接續或同時進行地,所述監控模組藉由所述第二攝影機產生對應至少一所述家禽的所述第二家禽影像。最後,所述雲端模組可依據所述第二家禽影像、所述至少一家禽影像特徵、以及所述影像重量關係式,而獲得各所述家禽的單位重量。或者,所述雲端模組可依據所述第二家禽影像、所述至少一家禽影像特徵,而獲得各所述家禽的活動力數值。進一步而言,所述學習校正模組可以隨時間不斷地進行重複機器學習之動作,以對於所述雲端模組所儲存之所述至少一家禽影像特徵以及所述影像重量關係式持續不斷地進行校正,以使本發明所述之家禽健康監測系統更加靈敏精確。由於前述學習、監測以及校正之動作均不需要多餘之人力介入,且可以無人職守地全時運作,不僅可節約人力成本,且不受時間早晚之限制,使得家禽飼養維護更具效率。 When using the poultry health monitoring system and method thereof in the present invention, the cloud module of the poultry health monitoring system may pre-store a weight judgment database. And, firstly, the artificial intelligence (artificial intelligence, AI) model is performed by the learning correction module. A machine learning (machine learning, ML) program, the learning correction module first senses the weight value of at least poultry carried by the weighing structure, and uses the first camera to generate the weight value of the weighing structure The first poultry image of the at least one poultry is carried. Finally, the learning correction module uses the calculation core to receive the weight value and the first poultry image to analyze the quantity of the at least one poultry in the first poultry image, and generate the at least one poultry image The features and the image weight relationship are stored in the cloud module, so as to complete the machine learning program. The cloud module is made to store at least a poultry image feature and an image weight relational expression. Continuing or simultaneously with respect to the above steps, the monitoring module generates the second poultry image corresponding to at least one poultry through the second camera. Finally, the cloud module can obtain the unit weight of each poultry according to the second poultry image, the feature of the at least one poultry image, and the image weight relational expression. Alternatively, the cloud module can obtain the activity value of each of the poultry according to the second poultry image and the features of the at least one poultry image. Further, the learning correction module can continuously repeat machine learning actions over time, so as to continuously perform the at least poultry image features and the image weight relationship stored in the cloud module. Correction, so that the poultry health monitoring system described in the present invention is more sensitive and accurate. Since the above-mentioned learning, monitoring and correction actions do not require redundant manpower intervention, and can be operated full-time without duty, it not only saves labor costs, but also is not limited by time, making poultry breeding and maintenance more efficient.

為此,本發明所述之家禽健康監測系統,可以解決現有技術之飼養維護成本難以降低、以及監控效率不彰之技術問題,達到低維護成本、可快速反應且全時監控之目的。 For this reason, the poultry health monitoring system described in the present invention can solve the technical problems of difficulty in reducing feeding and maintenance costs and poor monitoring efficiency in the prior art, and achieve the goals of low maintenance costs, quick response and full-time monitoring.

為了能更進一步瞭解本發明為達成預定目的所採取之技術、手段及功效,請參閱以下有關本發明之詳細說明與附圖,相信本發明 特徵與特點,當可由此得一深入且具體之瞭解,然而所附圖式僅提供參考與說明用,並非用來對本發明加以限制者。 In order to further understand the technology, means and effects that the present invention adopts to achieve the intended purpose, please refer to the following detailed description and accompanying drawings of the present invention, believe that the present invention The features and characteristics should be understood in depth and in detail, but the accompanying drawings are only for reference and illustration, and are not intended to limit the present invention.

10:雲端模組 10:Cloud module

11:伺服器 11:Server

12:雲端資料庫 12:Cloud database

20:學習校正模組 20: Learning Correction Module

21:秤重結構 21: weighing structure

22:第一攝影機 22: First camera

23:運算核心 23: Computing core

30:監控模組 30:Monitoring module

31:第二攝影機 31:Second camera

40:預警分析模組 40: Early warning analysis module

50:行動通訊平台 50:Mobile communication platform

100:雞隻 100: Chicken

210:重量感測器 210: weight sensor

211:秤重平台 211: weighing platform

212:中間平台 212: Middle platform

S1~S5:步驟 S1~S5: steps

圖1為本發明家禽健康監測系統之第一實施例的架構示意圖;圖2為本發明家禽健康監測系統之所述第一實施例的配置示意圖;圖3為本發明家禽健康監測系統之第二實施例的架構示意圖;以及圖4為本發明家禽健康監測方法之方法流程圖。 Fig. 1 is a schematic diagram of the structure of the first embodiment of the poultry health monitoring system of the present invention; Fig. 2 is a schematic configuration diagram of the first embodiment of the poultry health monitoring system of the present invention; Fig. 3 is a second embodiment of the poultry health monitoring system of the present invention A schematic diagram of the structure of the embodiment; and FIG. 4 is a flow chart of the poultry health monitoring method of the present invention.

以下係藉由特定的具體實施例說明本發明之實施方式,熟悉此技術之人士可由本說明書所揭示之內容輕易地瞭解本發明之其他優點及功效。本發明亦可藉由其他不同的具體實例加以施行或應用,本發明說明書中的各項細節亦可基於不同觀點與應用在不悖離本發明之精神下進行各種修飾與變更。 The implementation of the present invention is described below through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific examples, and various modifications and changes can be made to the details in the description of the present invention based on different viewpoints and applications without departing from the spirit of the present invention.

須知,本說明書所附圖式繪示之結構、比例、大小、元件數量等,均僅用以配合說明書所揭示之內容,以供熟悉此技術之人士瞭解與閱讀,並非用以限定本發明可實施之限定條件,故不具技術上之實質意義,任何結構之修飾、比例關係之改變或大小之調整,在不影響本發明所 能產生之功效及所能達成之目的下,均應落在本發明所揭示之技術內容得能涵蓋之範圍內。 It should be noted that the structure, proportion, size, number of components, etc. shown in the drawings attached to this specification are only used to match the content disclosed in the specification, for those who are familiar with this technology to understand and read, and are not used to limit the scope of the present invention. Implementation restrictions, so there is no technical substantive significance, any modification of the structure, the change of the proportional relationship or the adjustment of the size will not affect the present invention. The effects that can be produced and the goals that can be achieved should all fall within the scope covered by the technical content disclosed in the present invention.

茲有關本發明之技術內容及詳細說明,配合圖式說明如下。 Hereby, the technical content and detailed description of the present invention are described as follows in conjunction with the drawings.

請參閱圖1至圖2所示,其中,圖1為本發明家禽健康監測系統之第一實施例的架構示意圖;圖2為本發明家禽健康監測系統之所述第一實施例的配置示意圖。在本發明所述之第一實施例中,本發明所提出的所述家禽健康監測系統包括:雲端模組10、運算核心23、學習校正模組20以及監控模組30。其中,所述雲端模組10用以儲存至少一家禽影像特徵、對應各家禽影像特徵之影像重量關係式。在本發明所述之第一實施例中,所述雲端模組10包括一伺服器11以及一雲端資料庫12。其中所述伺服器11用以獲得各家禽之單位重量(即任一雞隻100之各別重量)、活動力之至少一者。並且,所述雲端資料庫12耦接伺服器11,且用以儲存至少一家禽影像特徵、影像重量關係式、單位重量、活動力之至少一者。進一步而言,所述伺服器11藉由窄頻物聯網(narrowband internet of things,NB-Iot)、LoRa WAN、LTE以及Wi-Fi之其中一者耦接所述雲端資料庫12。所述運算核心23耦接所述雲端模組10。 Please refer to FIG. 1 to FIG. 2 , wherein FIG. 1 is a schematic structural diagram of the first embodiment of the poultry health monitoring system of the present invention; FIG. 2 is a schematic configuration diagram of the first embodiment of the poultry health monitoring system of the present invention. In the first embodiment of the present invention, the poultry health monitoring system proposed by the present invention includes: a cloud module 10 , a computing core 23 , a learning correction module 20 and a monitoring module 30 . Wherein, the cloud module 10 is used to store at least one poultry image feature and an image weight relational expression corresponding to each poultry image feature. In the first embodiment of the present invention, the cloud module 10 includes a server 11 and a cloud database 12 . Wherein the server 11 is used to obtain at least one of the unit weight of each poultry (ie the individual weight of any chicken 100 ) and activity. Moreover, the cloud database 12 is coupled to the server 11 and is used to store at least one of poultry image features, image weight relational formula, unit weight, and activity. Further, the server 11 is coupled to the cloud database 12 through one of narrowband internet of things (NB-Iot), LoRa WAN, LTE and Wi-Fi. The computing core 23 is coupled to the cloud module 10 .

所述學習校正模組20耦接所述運算核心23以及所述雲端模組10,且所述學習校正模組20包括秤重結構21、第一攝影機22。其中,所述秤重結構21用以感測所承載之至少一家禽(如圖2所示之至少一雞隻100)的重量值。所述第一攝影機22配置於所述秤重結構21中,且用以產生所述秤重結構21所承載之所述至少一家禽的第一家禽影像。所述運算核心23配置於所述秤重結構21,所述運算核心23接收所述重量 值以及所述第一家禽影像,以分析所述第一家禽影像中之所述至少一家禽的數量,且產生所述至少一家禽影像特徵以及所述影像重量關係式。在本發明所述之第一實施例中,所述秤重結構21中包括一重量感測器210、一秤重平台211以及一中間平台212。其中,所述秤重平台211用以承載至少一家禽,所述中間平台212配置於所述秤重平台211之上,且所述第一攝影機22配置於所述中間平台212之下,用以俯拍至少一雞隻100。其中,所述秤重平台211藉由至少兩個柱體213耦接所述中間平台212,以使所述秤重平台211與所述中間平台212一起連動。在本發明所述之第一實施例中,所述運算核心23為利用一物件偵測演算工具作為所述運算核心23辨識目標物的深度學習架構,所述物件偵測演算工具為深度學習或影像處理方式例如包含至少一卷積層(convolution layer)以及至少一池化層(pooling layer)的一區域卷積神經網路(mask region-based convolutional neural networks,mask R-CNN)。所述雲端模組10藉由至少一卷積層與至少一池化層以比對第一家禽影像是否符合各家禽影像特徵,繼而依據影像重量關係式獲得單位重量(即任一雞隻100之各別重量)、或獲得活動力數值。如圖2所示,第一攝影機22可擷取至少一雞隻100(例如,如圖示之兩隻雞隻)之俯視影像,並將影像資料傳送至運算核心23,使運算核心23可計算出在秤重平台211上之雞隻數量。秤重平台211將在其上的雞隻總重量傳送至運算核心23,藉此運算核心23可計算出在秤重平台211上之雞隻平均重量。依據本發明一實施方式,若運算核心23判斷在秤重平台211僅有一隻雞隻,則可建立雞隻之俯視影像特徵(例如身長,俯視面積等)與重量之對應關係,亦即影像重量關係式(weight formula),然上述實施方式非限定本案範圍。例如運 算核心23也可計算秤重平台211上雞隻之俯視影像特徵平均值(例如平均身長)及重量平均值,也可以建立影像重量關係式。 The learning correction module 20 is coupled to the computing core 23 and the cloud module 10 , and the learning correction module 20 includes a weighing structure 21 and a first camera 22 . Wherein, the weighing structure 21 is used for sensing the weight value of at least one poultry (at least one poultry 100 shown in FIG. 2 ) carried. The first camera 22 is configured in the weighing structure 21 and is used to generate a first poultry image of the at least one poultry carried by the weighing structure 21 . The computing core 23 is configured in the weighing structure 21, and the computing core 23 receives the weight value and the first poultry image, to analyze the quantity of the at least one poultry in the first poultry image, and generate the at least one poultry image feature and the image weight relationship. In the first embodiment of the present invention, the weighing structure 21 includes a weight sensor 210 , a weighing platform 211 and an intermediate platform 212 . Wherein, the weighing platform 211 is used to carry at least one poultry, the intermediate platform 212 is arranged on the weighing platform 211, and the first camera 22 is arranged under the intermediate platform 212 for Overhead shot at least one chicken 100. Wherein, the weighing platform 211 is coupled to the middle platform 212 through at least two columns 213 , so that the weighing platform 211 and the middle platform 212 move together. In the first embodiment of the present invention, the computing core 23 is a deep learning framework using an object detection computing tool as the computing core 23 to identify objects, and the object detecting computing tool is deep learning or The image processing method is, for example, a mask region-based convolutional neural network (mask R-CNN) including at least one convolution layer and at least one pooling layer. The cloud module 10 uses at least one convolutional layer and at least one pooling layer to compare whether the first poultry image conforms to the characteristics of each poultry image, and then obtains the unit weight according to the image weight relationship (that is, each chicken 100 different weight), or obtain activity value. As shown in FIG. 2 , the first camera 22 can capture an overhead image of at least one chicken 100 (for example, two chickens as shown in the figure), and send the image data to the computing core 23, so that the computing core 23 can calculate The number of chickens on the weighing platform 211. The weighing platform 211 transmits the total weight of the chickens on it to the computing core 23, so that the computing core 23 can calculate the average weight of the chickens on the weighing platform 211. According to one embodiment of the present invention, if the calculation core 23 judges that there is only one chicken on the weighing platform 211, the corresponding relationship between the bird's bird's-eye view image features (such as body length, bird-view area, etc.) and weight can be established, that is, the image weight Relational formula (weight formula), but the above-mentioned implementation mode does not limit the scope of this case. For example shipped Calculation core 23 can also calculate the average value (such as average body length) and weight average value of the bird's-eye view image feature on the weighing platform 211, and also can establish the image weight relational expression.

依據本發明之一實施方式,運算核心23也可以藉由單一雞隻影像相對應於時間關係,判斷雞隻是否活動。例如若在日間運算核心23在一預定時間內(例如10分鐘)判斷一雞隻影像係為靜止或是沒有移動範圍在一預定數值之內(例如移動距離不超出1公尺),則可判斷該雞隻活動力不足。運算核心23可統合第一攝影機22及第二攝影機31之影像資料,並判斷雞隻中活動力不足之比例。若雞隻中活動力不足比例超出一門檻值,代表該養雞場之雞隻可能有傳染病發生,運算核心23可發出警示通知。 According to an embodiment of the present invention, the computing core 23 can also determine whether the chicken is moving or not by using a single chicken image corresponding to the time relationship. For example, if the daytime operation core 23 judges that a chicken image is still or does not have a moving range within a predetermined value (for example, the moving distance does not exceed 1 meter) within a predetermined time (for example, 10 minutes), then it can be determined The bird is not active enough. The computing core 23 can integrate the image data of the first camera 22 and the second camera 31, and judge the proportion of chickens with insufficient activity. If the lack of activity ratio of the chickens exceeds a threshold value, it means that the chickens in the chicken farm may have infectious diseases, and the calculation core 23 can issue a warning notice.

進一步而言,本發明所述之學習校正模組20針對紅羽土雞的機器學習之混淆矩陣法(Confusion Matrix)評估結果如下表:

Figure 110137175-A0305-02-0011-1
Further, the learning correction module 20 of the present invention is aimed at the evaluation results of the confusion matrix method (Confusion Matrix) of the machine learning of the red-feathered native chicken as follows:
Figure 110137175-A0305-02-0011-1

其中:真陽性(TP)意指人工計數「有」,深度學習計數「有」的個數;偽陽性(FP)意指人工計數「沒有」,深度學習計數「有」 的個數;偽陰性(FN)意指人工計數「有」,深度學習計數「沒有」的個數;及真陰性(TN)意指人工計數「沒有」,深度學習計數「沒有」的個數。前述結果為在三個月的飼養週期中,分別利用機器學習計數方式與人工計數方式而獲得。可證明本發明的機器學習計數可取代人工計數來有效評估雞隻平均重量。將每天收集到的多個平均重量的數據進一步換算成日標準差。無論是在紅羽公雞或紅羽母雞的實驗結果中,皆可觀察到飼養後期的雞隻平均體重的日標準差越來越大,推斷可能是因為成年的雞隻相較於幼年的雞隻較容易發生鬥爭,如此在進食時,瘦弱的雞隻無法與強壯的雞隻競爭食物,進而導致雞隻間顯著的體型差異。由此可知,平均重量的標準差有助於監控雞隻整體的健康狀態,若平均體重的標準差越來越大,則可考慮分區飼養以穩定雞隻的平均體重。 Among them: True Positive (TP) means the number of "Yes" counted manually and "Yes" counted by deep learning; False Positive (FP) means "No" counted manually and "Yes" counted by deep learning False Negative (FN) refers to the number of "Yes" counted manually and "No" counted by deep learning; and True Negative (TN) means the number of "No" counted manually and "No" counted by deep learning . The aforementioned results were obtained by using machine learning counting methods and manual counting methods respectively during the three-month feeding cycle. It can be proved that the machine learning counting of the present invention can replace manual counting to effectively estimate the average weight of chickens. The multiple average weight data collected every day were further converted into daily standard deviations. Whether it is in the experimental results of red-feathered cocks or red-feathered hens, it can be observed that the daily standard deviation of the average body weight of the chickens in the later rearing period is getting larger and larger. Fighting is prone to occur, so that when eating, thin birds cannot compete with stronger birds for food, resulting in significant size differences between birds. It can be seen that the standard deviation of the average weight helps to monitor the overall health status of the chickens. If the standard deviation of the average weight is getting larger and larger, you can consider breeding in different areas to stabilize the average weight of the chickens.

所述監控模組30耦接所述雲端模組10,且所述監控模組30包括第二攝影機31。進一步而言,所述監控模組30可利用先前建立之影像重量關係式快速判斷雞隻重量是否有異常狀況。其中,所述第二攝影機31用以產生包括至少一所述家禽的第二家禽影像(可以是承載於秤重結構21之外的任何其他雞隻100)。或者,所述雲端模組10依據所述第二家禽影像、所述至少一家禽影像特徵,而獲得各所述家禽的活動力數值。進一步而言,所述活動力數值,可以是所述雲端模組10依據所述第二家禽影像中之各別雞隻100的活動距離、活動頻率、靜止週期時間等判斷條件,來作為雲端模組10判斷所述活動力數值之依據,例如活動距離短且活動頻率低則判斷為活動力不佳,可更設定有閥值來作為群體分類。 The monitoring module 30 is coupled to the cloud module 10 , and the monitoring module 30 includes a second camera 31 . Furthermore, the monitoring module 30 can quickly determine whether the weight of the chicken is abnormal by using the previously established image weight relational expression. Wherein, the second camera 31 is used to generate a second poultry image including at least one poultry (it may be any other chicken 100 carried outside the weighing structure 21 ). Alternatively, the cloud module 10 obtains the activity value of each poultry according to the second poultry image and the features of the at least one poultry image. Further, the activity value may be determined by the cloud module 10 as a cloud model according to the judgment conditions of the individual chickens 100 in the second poultry image, such as the activity distance, activity frequency, and resting cycle time. The basis for group 10 to judge the activity value, for example, if the activity distance is short and the activity frequency is low, it is judged as poor activity, and a threshold value can be set as a group classification.

請參閱圖3所示,為本發明家禽健康監測系統之第二實施例的架構示意圖。在本發明所述之第二實施例中,其與前述第一實施例大致相同,惟更包括一預警分析模組40、一行動通訊平台50。其中,所述預警分析模組40耦接所述雲端模組10,且所述預警分析模組40依據單位重量、活動力之至少一者,以輸出一統計報告以及一警示訊息之至少一者。所述行動通訊平台50無線地耦接所述預警分析模組40,且接收所述統計報告以及所述警示訊息之至少一者,以供家禽之飼養業者可提前預知雞隻100之健康狀況或者雞隻100的生長趨勢,使得家禽飼養業者可以提早進行應變或提早採取防範措施,以降低飼養家禽之風險以及成本。在本發明之所述第二實施例中,所述行動通訊平台50包括工作站、伺服器、桌上型電腦、筆記型電腦、平板電腦、個人數位助理或智慧型手機的其中一者。然而,本發明不受此限制。 Please refer to FIG. 3 , which is a schematic structural diagram of the second embodiment of the poultry health monitoring system of the present invention. In the second embodiment of the present invention, it is substantially the same as the aforementioned first embodiment, but further includes an early warning analysis module 40 and a mobile communication platform 50 . Wherein, the early warning analysis module 40 is coupled to the cloud module 10, and the early warning analysis module 40 outputs at least one of a statistical report and a warning message according to at least one of unit weight and activity force . The mobile communication platform 50 is wirelessly coupled to the early warning analysis module 40, and receives at least one of the statistical report and the warning message, so that the poultry breeder can predict the health status of the chicken 100 in advance or The growth trend of chickens 100 enables poultry farmers to respond early or take preventive measures in advance to reduce the risk and cost of raising poultry. In the second embodiment of the present invention, the mobile communication platform 50 includes one of a workstation, a server, a desktop computer, a notebook computer, a tablet computer, a personal digital assistant or a smart phone. However, the present invention is not limited thereto.

請參閱圖4所示,為本發明家禽健康監測方法之方法流程圖。並且,首先由所述學習校正模組20先進行人工智慧(artificial intelligence,AI)模型之機器學習(machine learning,ML)程序,所述學習校正模組20先以所述秤重結構21感測所承載之至少一家禽的重量值,且以所述第一攝影機22產生所述秤重結構21所承載之所述至少一家禽的所述第一家禽影像。最後學習校正模組再以所述運算核心23接收所述重量值以及所述第一家禽影像,以分析所述第一家禽影像中之所述至少一家禽的數量,且產生所述至少一家禽影像特徵以及所述影像重量關係式以儲存於所述雲端模組(步驟S1),藉此完成所述機器學習程序。使得所述雲端模組10儲存至少一家禽影像特徵、影像重量關係式(步驟S2)。相對於前述步驟可接續或同時進行地,所述監控模組30藉由所述 第二攝影機31產生對應至少一所述家禽的所述第二家禽影像(步驟S3)。最後,所述雲端模組10可依據所述第二家禽影像、所述至少一家禽影像特徵、以及所述影像重量關係式,而獲得各所述家禽的單位重量(步驟S4)。或者,所述雲端模組10可依據所述第二家禽影像、所述至少一家禽影像特徵,而獲得各所述家禽的活動力數值(步驟S5)。進一步而言,所述學習校正模組20可以隨時間不斷地進行重複機器學習之動作,以對於所述雲端模組10所儲存之所述至少一家禽影像特徵以及所述影像重量關係式持續不斷地進行校正,以使本發明所述之家禽健康監測系統更加靈敏精確。由於前述學習、監測以及校正之動作均不需要多餘之人力介入,且可以無人職守地全時運作,不僅可節約人力成本,且不受時間早晚之限制,使得家禽飼養維護更具效率。 Please refer to FIG. 4 , which is a flow chart of the poultry health monitoring method of the present invention. And, firstly, the learning correction module 20 performs the machine learning (machine learning, ML) program of the artificial intelligence (artificial intelligence, AI) model, and the learning correction module 20 uses the weighing structure 21 to sense The weight value of the at least one poultry carried, and the first poultry image of the at least one poultry carried by the weighing structure 21 is generated by the first camera 22 . Finally, the learning correction module uses the calculation core 23 to receive the weight value and the first poultry image to analyze the quantity of the at least one poultry in the first poultry image, and generate the at least one poultry Image features and the image weight relationship are stored in the cloud module (step S1), thereby completing the machine learning program. Make the cloud module 10 store at least one poultry image feature and image weight relationship (step S2). With respect to the foregoing steps, which may be carried out sequentially or simultaneously, the monitoring module 30 uses the The second camera 31 generates the second poultry image corresponding to at least one poultry (step S3). Finally, the cloud module 10 can obtain the unit weight of each poultry according to the second poultry image, the feature of the at least one poultry image, and the image weight relational expression (step S4). Alternatively, the cloud module 10 can obtain the activity value of each poultry according to the second poultry image and the features of the at least one poultry image (step S5). Further, the learning correction module 20 can continuously repeat the machine learning action over time, so that the at least poultry image features and the image weight relationship stored in the cloud module 10 can be continuously Calibration is carried out accurately, so that the poultry health monitoring system described in the present invention is more sensitive and accurate. Since the above-mentioned learning, monitoring and correction actions do not require redundant manpower intervention, and can be operated full-time without duty, it not only saves labor costs, but also is not limited by time, making poultry breeding and maintenance more efficient.

為此,本發明所述之家禽健康監測系統,可以解決現有技術之飼養維護成本難以降低、以及監控效率不彰之技術問題,達到低維護成本、可快速反應且全時監控之目的。 For this reason, the poultry health monitoring system described in the present invention can solve the technical problems of difficulty in reducing feeding and maintenance costs and poor monitoring efficiency in the prior art, and achieve the goals of low maintenance costs, quick response and full-time monitoring.

以上所述,僅為本發明較佳具體實施例之詳細說明與圖式,惟本發明之特徵並不侷限於此,並非用以限制本發明,本發明之所有範圍應以下述之申請專利範圍為準,凡合於本發明申請專利範圍之精神與其類似變化之實施例,皆應包括於本發明之範疇中,任何熟悉該項技藝者在本發明之領域內,可輕易思及之變化或修飾皆可涵蓋在以下本案之專利範圍。 The above is only a detailed description and drawings of preferred embodiments of the present invention, but the features of the present invention are not limited thereto, and are not intended to limit the present invention. As the standard, all embodiments that conform to the spirit of the patent scope of the present invention and its similar changes should be included in the scope of the present invention. Any person familiar with the art can easily think of changes or Modifications can all be covered by the patent scope of the following case.

10:雲端模組 10:Cloud module

11:伺服器 11:Server

12:雲端資料庫 12:Cloud database

20:學習校正模組 20: Learning Correction Module

21:秤重結構 21: weighing structure

22:第一攝影機 22: First camera

23:運算核心 23: Computing core

30:監控模組 30:Monitoring module

31:第二攝影機 31:Second camera

Claims (8)

一種家禽健康監測系統,包括:一雲端模組,用以儲存至少一家禽影像特徵、以及對應各家禽影像特徵之一影像重量關係式;一運算核心,耦接該雲端模組,該運算核心接收一重量值以及一第一家禽影像,以分析該第一家禽影像中之該至少一家禽的數量,且產生該至少一家禽影像特徵以及該影像重量關係式,且該影像重量關係式包括影像特徵值與重量之相對關係;一學習校正模組,耦接該運算核心以及該雲端模組,且包括一秤重結構、一第一攝影機;其中,該秤重結構用以感測所承載之至少一家禽的重量值;該第一攝影機配置於該秤重結構中,且用以產生該秤重結構所承載之該至少一家禽的一第一家禽影像;以及一監控模組,耦接該雲端模組,且包括一第二攝影機;其中,該第二攝影機用以產生包括至少一該家禽的一第二家禽影像;一預警分析模組,該預警分析模組耦接該雲端模組,且該預警分析模組依據各該家禽的一單位重量以及該各該家禽的一活動力之至少一者,以輸出一統計報告以及一警示訊息之至少一者;其中,該雲端模組依據該第二家禽影像、該至少一家禽影像特徵、以及該影像重量關係式,而獲得該單位重量。 A poultry health monitoring system, comprising: a cloud module for storing at least one poultry image feature and an image weight relational expression corresponding to each poultry image feature; a computing core coupled to the cloud module, the computing core receiving A weight value and a first poultry image, to analyze the quantity of the at least one poultry in the first poultry image, and generate the at least one poultry image feature and the image weight relationship, and the image weight relationship includes the image feature The relative relationship between value and weight; a learning correction module, coupled to the computing core and the cloud module, and includes a weighing structure and a first camera; wherein, the weighing structure is used to sense at least The weight value of a poultry; the first camera is configured in the weighing structure, and is used to generate a first poultry image of the at least one poultry carried by the weighing structure; and a monitoring module, coupled to the cloud module, and includes a second camera; wherein, the second camera is used to generate a second poultry image including at least one poultry; an early warning analysis module, the early warning analysis module is coupled to the cloud module, and The early warning analysis module outputs at least one of a statistical report and a warning message based on at least one of a unit weight of each poultry and an activity of each poultry; Two poultry images, the at least one poultry image feature, and the image weight relationship formula to obtain the unit weight. 如請求項1所述之家禽健康監測系統,更包括一行動通訊平台,該行動通訊平台無線地耦接該預警分析模組,且接收該統計報告以及該警示訊息之至少一者。 The poultry health monitoring system as described in Claim 1 further includes a mobile communication platform, which is wirelessly coupled to the early warning analysis module, and receives at least one of the statistical report and the warning message. 如請求項2所述之家禽健康監測系統,其中,該行動通訊平台包括工作站、伺服器、桌上型電腦、筆記型電腦、平板電腦、個人數位助理或智慧型手機的其中一者。 The poultry health monitoring system as described in Claim 2, wherein the mobile communication platform includes one of a workstation, a server, a desktop computer, a notebook computer, a tablet computer, a personal digital assistant, or a smart phone. 如請求項1所述之家禽健康監測系統,其中,該運算核心為利用一物件偵測演算工具作為該運算核心辨識目標物的深度學習架構,所述物件偵測演算工具為深度學習或影像處理方式,該雲端模組藉由至少一卷積層與至少一池化層以比對該第二家禽影像是否符合各該家禽影像特徵,繼而獲得該單位重量、活動力數值。 The poultry health monitoring system as described in Claim 1, wherein, the calculation core is a deep learning framework that uses an object detection calculation tool as the calculation core to identify objects, and the object detection calculation tool is deep learning or image processing In this way, the cloud module uses at least one convolutional layer and at least one pooling layer to compare whether the second poultry image conforms to the characteristics of each poultry image, and then obtain the unit weight and activity values. 如請求項1所述之家禽健康監測系統,其中,該雲端模組包括一伺服器以及一雲端資料庫;其中該伺服器用以獲得該單位重量、各該家禽的一活動力之至少一者;該雲端資料庫耦接該伺服器,且用以儲存該至少一家禽影像特徵、該影像重量關係式、該單位重量、該活動力之至少一者。 The poultry health monitoring system as described in claim 1, wherein the cloud module includes a server and a cloud database; wherein the server is used to obtain at least one of the unit weight and an activity of each poultry ; The cloud database is coupled to the server, and is used to store at least one of the at least one poultry image feature, the image weight relationship, the unit weight, and the activity force. 如請求項5所述之家禽健康監測系統,其中,該伺服器藉由窄頻物聯網(narrowband internet of things,NB-Iot)、LoRa WAN、LTE以及Wi-Fi之其中一者耦接該雲端資料庫。 The poultry health monitoring system as described in claim 5, wherein the server is coupled to the cloud through one of narrowband internet of things (NB-Iot), LoRa WAN, LTE and Wi-Fi database. 如請求項1所述之家禽健康監測系統,其中,該秤重結構中包括一秤重平台以及一中間平台;其中該秤重平台用以承載至少一該家禽,該中間平台配置於該秤重平台之上,且該第一攝影機配置於該中間平台之下。 The poultry health monitoring system as described in claim 1, wherein the weighing structure includes a weighing platform and an intermediate platform; wherein the weighing platform is used to carry at least one poultry, and the intermediate platform is configured on the weighing platform above the platform, and the first camera is disposed under the intermediate platform. 如請求項7所述之家禽健康監測系統,其中,該秤重平台藉由至少兩個柱體耦接該中間平台。 The poultry health monitoring system according to claim 7, wherein the weighing platform is coupled to the intermediate platform by at least two columns.
TW110137175A 2021-10-06 2021-10-06 Poultry health monitoring system and method thereof TWI784740B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW110137175A TWI784740B (en) 2021-10-06 2021-10-06 Poultry health monitoring system and method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW110137175A TWI784740B (en) 2021-10-06 2021-10-06 Poultry health monitoring system and method thereof

Publications (2)

Publication Number Publication Date
TWI784740B true TWI784740B (en) 2022-11-21
TW202315518A TW202315518A (en) 2023-04-16

Family

ID=85794625

Family Applications (1)

Application Number Title Priority Date Filing Date
TW110137175A TWI784740B (en) 2021-10-06 2021-10-06 Poultry health monitoring system and method thereof

Country Status (1)

Country Link
TW (1) TWI784740B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023061375A1 (en) * 2021-10-15 2023-04-20 智逐科技股份有限公司 Domestic fowl health monitoring system and method

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI740672B (en) * 2020-10-07 2021-09-21 國立成功大學 System and method for smart aquaculture

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI740672B (en) * 2020-10-07 2021-09-21 國立成功大學 System and method for smart aquaculture

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023061375A1 (en) * 2021-10-15 2023-04-20 智逐科技股份有限公司 Domestic fowl health monitoring system and method

Also Published As

Publication number Publication date
TW202315518A (en) 2023-04-16

Similar Documents

Publication Publication Date Title
AU2017276810B2 (en) System for monitoring pasture intake
Long et al. Linking habitat selection to fitness-related traits in herbivores: the role of the energy landscape
WO2020133560A1 (en) Big data technology-based intelligent livestock breeding management system and method
CN107157458A (en) The sensor-based system and method for a kind of animal individual feed intake and health monitoring
JP6781440B1 (en) Livestock information management system, livestock information management server, livestock information management method, and livestock information management program
US9961883B2 (en) Rapid and automatic determination of metabolic efficiency in livestock
WO2021203757A1 (en) Livestock and poultry breeding method for individual control
TWI784740B (en) Poultry health monitoring system and method thereof
Patel et al. Role of artificial intelligence in livestock and poultry farming
CN107168403B (en) A kind of environment of chicken house system for detecting temperature based on CAN fieldbus
Maltz Novel technologies: sensors, data and precision dairy farming
WO2023061375A1 (en) Domestic fowl health monitoring system and method
CN113516139A (en) Data processing method, device, equipment and storage medium
CN111831038A (en) Modern pig raising daily intelligent management system based on intelligent control
Ojukwu et al. Development of a computer vision system to detect inactivity in group-housed pigs
Shen et al. A review of information perception technology in livestock breeding.
Bello et al. A framework for real-time cattle monitoring using multimedia networks
CN110779575A (en) Animal husbandry monitoring system based on big data platform
CN115777560A (en) Intelligent sow feeding system based on machine vision analysis technology
CN107714000B (en) Milk cow health condition detection method and device
Fu et al. Research and application of a new multilevel fuzzy comprehensive evaluation method for cold stress in dairy cows
Salau et al. How does the behaviour of dairy cows during recording affect an image processing based calculation of the udder depth?
Ronghua et al. Cow behavioral recognition using dynamic analysis
KR20210051860A (en) Method for providing management guidline based on feeding standard
Azarpajouh et al. Smart farming: A review of animal-based measuring technologies for broiler welfare assessment